Skip to content

Commit

Permalink
Misc CI/CD updates (#240)
Browse files Browse the repository at this point in the history
  • Loading branch information
dbarbuzzi committed May 14, 2024
1 parent d485d3e commit 3a25456
Show file tree
Hide file tree
Showing 6 changed files with 19 additions and 18 deletions.
6 changes: 3 additions & 3 deletions .github/workflows/build-test.yml
Original file line number Diff line number Diff line change
Expand Up @@ -139,7 +139,7 @@ jobs:

TEST-MULTI:
needs: [BUILD]
if: success() && contains(fromJSON('["NIGHTLY", "RELEASE"]'), inputs.wf_category)
if: success() && contains(fromJSON('["NIGHTLY", "WEEKLY", "RELEASE"]'), inputs.wf_category)
uses: ./.github/workflows/test.yml
with:
test_label: ${{ inputs.test_label_multi }}
Expand Down Expand Up @@ -180,7 +180,7 @@ jobs:
TEST-ACCURACY-SMOKE:
needs: [BUILD]
if: inputs.wf_category == 'NIGHTLY'
uses: ./.github/workflows/nm-lm-eval-smoke.yml
uses: ./.github/workflows/nm-test-accuracy-smoke.yml
with:
label: ${{ inputs.test_label_solo }}
timeout: ${{ inputs.benchmark_timeout }}
Expand All @@ -194,7 +194,7 @@ jobs:
TEST-ACCURACY-FULL:
needs: [BUILD]
if: ${{ inputs.wf_category == 'WEEKLY' || inputs.wf_category == 'RELEASE' }}
uses: ./.github/workflows/nm-lm-eval-accuracy.yml
uses: ./.github/workflows/nm-test-accuracy-full.yml
with:
label: ${{ inputs.test_label_multi }}
timeout: ${{ inputs.benchmark_timeout }}
Expand Down
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
name: nm-lm-eval-accuracy
name: nm-test-accuracy-full
on:
# makes workflow reusable
workflow_call:
Expand Down Expand Up @@ -68,7 +68,7 @@ env:
VENV_BASE: "LM_EVAL"

jobs:
LM-EVAL-FULL:
TEST-ACCURACY-FULL:

runs-on: ${{ inputs.label }}
timeout-minutes: ${{ fromJSON(inputs.timeout) }}
Expand Down
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
name: nm-lm-eval-smoke
name: nm-test-accuracy-smoke
on:
# makes workflow reusable
workflow_call:
Expand Down Expand Up @@ -68,7 +68,7 @@ env:
VENV_BASE: "LM_EVAL"

jobs:
LM-EVAL-SMOKE:
TEST-ACCURACY-SMOKE:

runs-on: ${{ inputs.label }}
timeout-minutes: ${{ fromJSON(inputs.timeout) }}
Expand Down
2 changes: 1 addition & 1 deletion neuralmagic/benchmarks/run_benchmark_serving.py
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ def get_tensor_parallel_size(config: NamedTuple) -> int:
return tensor_parallel_size


def is_server_running(host: str, port: int, timeout=300) -> bool:
def is_server_running(host: str, port: int, timeout=600) -> bool:

def try_connection() -> bool:
try:
Expand Down
19 changes: 10 additions & 9 deletions tests/accuracy/lm-eval-tasks.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -63,12 +63,13 @@
# value: 0.5041698256254739

# Mixtral: FP16
- model_name: "mistralai/Mixtral-8x7B-Instruct-v0.1"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.6550416982562547
- name: "exact_match,flexible-extract"
value: 0.6603487490523123
enable_tensor_parallel: true
# g5.12xlarge runner (4x 24GB A10 GPUs) has insufficient VRAM
# - model_name: "mistralai/Mixtral-8x7B-Instruct-v0.1"
# tasks:
# - name: "gsm8k"
# metrics:
# - name: "exact_match,strict-match"
# value: 0.6550416982562547
# - name: "exact_match,flexible-extract"
# value: 0.6603487490523123
# enable_tensor_parallel: true
2 changes: 1 addition & 1 deletion tests/utils/server.py
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@

from tests.utils.logging import log_banner

MAX_SERVER_START_WAIT = 600 # time (seconds) to wait for server to start
MAX_SERVER_START_WAIT = 15 * 60 # time (seconds) to wait for server to start


@ray.remote(num_gpus=torch.cuda.device_count())
Expand Down

5 comments on commit 3a25456

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: 3a25456 Previous: d485d3e Ratio
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.424070416407297 prompts/s 17.45885380130362 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2265.1291541329488 tokens/s 2269.6509941694703 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.007501906142668 prompts/s 2.0090910889604428 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4113.371405686327 tokens/s 4116.627641279947 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0423199908810425 prompts/s 2.0420892274236246 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4184.713661315256 tokens/s 4184.2408269910065 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.341585871253171 prompts/s 10.366746516695056 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1344.4061632629123 tokens/s 1347.6770471703574 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2407312467952546 prompts/s 0.2408087883142535 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.295062083383097 tokens/s 31.305142480852957 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.366402860785677 prompts/s 5.374808985672395 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 697.6323719021381 tokens/s 698.7251681374114 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.656469648974914 prompts/s 5.653633475977627 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2644.738949074711 tokens/s 2643.4128680280996 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.816900738459765 prompts/s 16.82528216827471 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2186.1970959997698 tokens/s 2187.2866818757125 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.943917980271479 prompts/s 13.959139226272045 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3583.58692092977 tokens/s 3587.4987811519154 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.232159077584827 prompts/s 14.267524211368551 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3657.6648829393002 tokens/s 3666.7537223217178 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7657654287399502 prompts/s 3.7626156458153983 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.909564458449 tokens/s 3856.6810369607833 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.1220280788311 prompts/s 8.11569046610624 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4166.600404440354 tokens/s 4163.349209112502 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9838164248355225 prompts/s 0.9838550931346398 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.0815522508366 tokens/s 290.09295372772567 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.34693713649918 tokens/s 212.35528330218065 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9133609973407109 prompts/s 0.9137459158052325 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.73692965429241 tokens/s 118.78696905468021 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7199570297394229 prompts/s 0.7234133412978745 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 93.59441386612498 tokens/s 94.04373436872369 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49188704204744005 prompts/s 0.4919132458930308 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.0254206948203 tokens/s 130.03234741936376 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.98142793387117 tokens/s 121.99448498147164 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.440966165357702 prompts/s 6.457116453290042 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 837.3256014965013 tokens/s 839.4251389277055 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.28064886662881 prompts/s 20.247217206011417 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2636.484352661745 tokens/s 2632.138236781484 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.578033470060364 prompts/s 11.583829296115017 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1505.1443511078471 tokens/s 1505.8978084949522 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3002259800278337 prompts/s 2.300974144441723 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 710.3711219920625 tokens/s 710.6021751141226 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 460.30895525127664 tokens/s 460.6703635448626 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 22.899320002747018 prompts/s 24.443938891144068 prompts/s 1.07
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2954.0122803543654 tokens/s 3153.268116957585 tokens/s 1.07
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.2020350880334005 prompts/s 3.2034136753545686 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1509.330471305952 tokens/s 1509.9802904971816 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.584089335905071 prompts/s 11.605228976145973 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1505.9316136676591 tokens/s 1508.6797668989766 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2564869094961478 prompts/s 0.25672487636671604 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.34329823449921 tokens/s 33.37423392767309 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.103219281453608 prompts/s 26.027830568603793 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1696.7092532944844 tokens/s 1691.8089869592466 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.3554262125099568 prompts/s 2.3557112612966504 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 727.4184257887416 tokens/s 727.5064564553736 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 540.5483317930516 tokens/s 540.6137478291425 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.553026266760048 prompts/s 6.5709542389398585 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 851.8934146788063 tokens/s 854.2240510621816 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.483700126985231 prompts/s 7.486450852726426 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3839.1381651434235 tokens/s 3840.5492874486563 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.564270518952236 prompts/s 11.588339878756337 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1503.3551674637908 tokens/s 1506.4841842383237 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0628492488874315 prompts/s 2.0644265494330707 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4226.7781109703465 tokens/s 4230.009999788362 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.729472655506023 prompts/s 6.735574667216951 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 874.831445215783 tokens/s 875.6247067382036 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.24045908874356436 prompts/s 0.24059288698853612 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.25968153666337 tokens/s 31.277075308509698 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.7312810751164855 prompts/s 6.748043912141075 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 875.066539765143 tokens/s 877.2457085783398 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.800970162049864 prompts/s 1.8008286607629982 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.1261210664823 tokens/s 234.10772589918977 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.9831215100438788 prompts/s 1.985354238826718 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 257.80579630570423 tokens/s 258.09605104747334 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.044172934692067 prompts/s 8.040437806110551 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4126.660715497031 tokens/s 4124.7445945347135 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7579242189771024 prompts/s 0.7581851684336032 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 98.53014846702331 tokens/s 98.56407189636842 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.396518202442901 prompts/s 3.403168528141097 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.54736631757714 tokens/s 442.4119086583426 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.7981103731056467 prompts/s 2.613947703296607 prompts/s 0.93
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 363.75434850373404 tokens/s 339.81320142855895 tokens/s 0.93
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.800347262321508 prompts/s 1.8018837875142801 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.045144101796 tokens/s 234.24489237685643 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.4466006598291634 prompts/s 2.4469783979387754 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.5755264395077 tokens/s 755.6921820407722 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.2632399016625 tokens/s 560.9322773920293 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.655907339372177 prompts/s 5.656258130386432 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2644.476035596855 tokens/s 2644.64005144348 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.439933266653582 prompts/s 4.44196346527194 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1406.9704528698535 tokens/s 1407.6138025100254 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1014.8947458692312 tokens/s 1015.4624612509341 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46397536248664073 prompts/s 0.463520024094734 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.6472473197186 tokens/s 122.52688316920198 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.07826257301987 tokens/s 114.96532650946989 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.20018458056593 prompts/s 49.29705602884645 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3198.011997736786 tokens/s 3204.308641875019 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.622100497857929 prompts/s 5.624589197137983 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 730.8730647215307 tokens/s 731.1965956279378 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.015109728224442 prompts/s 7.042005918148534 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3279.98470452862 tokens/s 3292.560287089528 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.167273960013263 prompts/s 4.17260055673134 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4271.455809013595 tokens/s 4276.915570649624 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.914449553756308 prompts/s 0.9143859155159352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.87844198832003 tokens/s 118.87016901707158 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.47501350599298997 prompts/s 0.475098682045607 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.56507017418697 tokens/s 125.58758561193576 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.4361605878659 tokens/s 99.4539907748804 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8369140082775717 prompts/s 3.8408865302828166 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1473.3749791785876 tokens/s 1474.9004276286016 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 47.98356045005688 prompts/s 48.146230502846066 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3118.9314292536974 tokens/s 3129.5049826849945 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.376262180247515 prompts/s 4.376621131542948 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1386.793722298635 tokens/s 1386.9074703746448 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1000.2238963767712 tokens/s 999.9733138823539 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.451743468411279 prompts/s 2.4517680963310378 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 757.1637628712273 tokens/s 757.1713686295933 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 563.3746901367086 tokens/s 562.997873448432 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.17391754074011 prompts/s 4.17446080449878 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1322.6727294851335 tokens/s 1322.8448843376184 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 833.4283762531284 tokens/s 833.4895420694445 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.943105492581521 prompts/s 13.935092429466788 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3583.3781115934507 tokens/s 3581.3187543729646 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0161613181944231 prompts/s 1.0166856525214367 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.100971365275 tokens/s 132.16913482778676 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.9259224235950896 prompts/s 2.9263246157622866 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 380.3699150673616 tokens/s 380.4222000490973 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9476499578299636 prompts/s 0.947734868686396 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.41774889935755 tokens/s 279.44278514841284 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.56287873036874 tokens/s 204.5812078708744 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.502566656046877 prompts/s 24.43199699436081 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3160.831098630047 tokens/s 3151.7276122725443 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839548629424585 prompts/s 0.983977399518168 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.12237118812703 tokens/s 290.1290161725969 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 211.8126834960799 tokens/s 212.2144057540833 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.24043629953692333 prompts/s 0.2406090203706557 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.25671893980003 tokens/s 31.27917264818524 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9144479176987446 prompts/s 0.9141892722282193 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.8782293008368 tokens/s 118.84460538966852 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.04326448262223 prompts/s 2.041595395165156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4186.648924892949 tokens/s 4183.228964693405 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.929259558681293 prompts/s 13.969617172337554 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3579.8197065810923 tokens/s 3590.1916132907513 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.596594944596731 prompts/s 15.578473913603274 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4008.32490076136 tokens/s 4003.667795796041 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.143693794019235 prompts/s 26.034278082038465 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1699.3400966112504 tokens/s 1692.2280753325003 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4878618159519563 prompts/s 3.4777941152484564 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1105.2685308570153 tokens/s 1102.0781771810832 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 797.3205606441959 tokens/s 794.9680900399532 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.556160185117204 prompts/s 25.49805802814238 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1661.1504120326183 tokens/s 1657.3737718292548 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.488704618486003 prompts/s 7.4912303349518945 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3841.7054692833194 tokens/s 3843.001161830322 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0716424679274215 prompts/s 2.0726681559409985 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4244.795416783287 tokens/s 4246.897051523106 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.96805363714151 prompts/s 0.9681604070981638 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.6510066773983 tokens/s 288.68284298717344 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 210.04182449605057 tokens/s 210.06499072944558 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.7931292089984 prompts/s 24.186631400191875 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3069.3136679607937 tokens/s 3120.0754506247517 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7651178234648914 prompts/s 3.7646441882254944 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.245769051514 tokens/s 3858.760292931132 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.765628218519809 prompts/s 3.764889148332317 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.768923982804 tokens/s 3859.011377040625 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.663133950928195 prompts/s 20.702290603550058 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2686.207413620665 tokens/s 2691.2977784615077 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3962959507302286 prompts/s 3.399009142352374 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.51847359492973 tokens/s 441.8711885058086 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0418823808224134 prompts/s 2.0417904814497367 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4183.816998305125 tokens/s 4183.628696490511 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.727147441265542 prompts/s 3.7273345040302686 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 484.52916736452045 tokens/s 484.55348552393497 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4661523237545599 prompts/s 0.46622566694755396 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.05380370754972 tokens/s 126.07363668484456 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.11989115724711 tokens/s 118.1384758333972 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.891541247086469 prompts/s 9.893361252236701 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1285.900362121241 tokens/s 1286.1369627907711 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.223504649683515 prompts/s 2.228676699666547 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 686.6775292795943 tokens/s 688.2747962356874 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 510.4217980356152 tokens/s 511.6061084686538 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.58705916019339 prompts/s 28.675999289815444 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3687.7306316649474 tokens/s 3699.2039083861923 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.283460135276675 prompts/s 10.296198262914897 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1336.8498175859677 tokens/s 1338.5057741789367 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.49192321536506084 prompts/s 0.4918963276896557 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03498274960018 tokens/s 130.0278752614836 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.00023689863752 tokens/s 121.99356857588587 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839197953119787 prompts/s 0.9839203228160437 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.112031380388 tokens/s 290.11218691671985 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.60538937101236 tokens/s 212.38904088307117 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8724458620169036 prompts/s 3.872757050662234 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3969.257008567326 tokens/s 3969.5759769287897 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.658118828175657 prompts/s 5.651181668821501 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2645.51003930181 tokens/s 2642.2665010741807 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.02224688684285 prompts/s 26.106528472473716 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1691.4460476447853 tokens/s 1696.9243507107915 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.229966994538023 prompts/s 7.2590351137115885 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3380.443367966198 tokens/s 3394.0344577669903 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.0532619042704 prompts/s 4.0550726389644725 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4154.59345187716 tokens/s 4156.4494549385845 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.830150158156309 prompts/s 1.8273191772228683 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 581.4948298511096 tokens/s 580.5953403918068 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 409.7852616124629 tokens/s 409.15625518475724 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.107892858450175 prompts/s 29.174460501148605 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3754.9181787400726 tokens/s 3763.50540464817 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8001479114482124 prompts/s 1.8026610572464896 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.0192284882676 tokens/s 234.34593744204363 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.393419741477225 prompts/s 3.3993142849262794 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.1445663920393 tokens/s 441.9108570404164 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.889925441151811 prompts/s 7.891096679323421 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4047.531751310879 tokens/s 4048.1325964929147 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.939859039356216 prompts/s 15.931149649188745 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4096.543773114548 tokens/s 4094.3054598415074 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.695894085629463 prompts/s 1.7036629049511658 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 553.0943747029647 tokens/s 555.6280767196934 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 385.46089731877777 tokens/s 387.23576364378016 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.487194678544224 prompts/s 24.499635371471232 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3158.848113532205 tokens/s 3160.4529629197887 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.5247466852304163 prompts/s 3.5137946523034143 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1116.9569770826668 tokens/s 1113.486387368429 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 805.780590554908 tokens/s 803.2862529333154 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.486571243555397 prompts/s 7.491801833614156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3840.6110479439185 tokens/s 3843.2943406440622 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: 3a25456 Previous: d485d3e Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3647.388925000996 ms 3631.2452459997075 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 142.28233248331284 ms 143.76172412334503 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 94.0930034998928 ms 95.35134399993694 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.38069158768323 ms 24.310515078398957 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.72852694291979 ms 23.651293949528196 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2035.789914500583 ms 2033.819562499957 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.91660104005011 ms 82.19037280665968 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.09485599999607 ms 38.89878250038237 ms 1.03
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.626430329528024 ms 11.560121123356186 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.73498152922247 ms 11.691221148500436 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5864.5879960004095 ms 5856.753418500375 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.7299674492994 ms 118.7610248973045 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.89390899983118 ms 89.27816299910774 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 53.37906246921517 ms 53.309391464699814 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 46.33974202275668 ms 46.510393959855044 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5303.213345499898 ms 5320.154192500013 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 181.10532584798784 ms 181.2350845293695 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 174.4023470000684 ms 171.38535149933887 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.6689300470948 ms 39.67783502514967 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.83108029045941 ms 36.79312972752162 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2574.683739000193 ms 2571.6090910000275 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 120.37171372000618 ms 120.15518762534703 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.68219500016494 ms 81.6943180002454 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.21018293462904 ms 19.216760619551575 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.287023850474057 ms 17.232633518539803 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13713.915370999985 ms 13589.161211999908 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 199.60011985066558 ms 203.4137675746612 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 161.19895849988097 ms 159.45165350012758 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 106.98148762674812 ms 106.80662248274147 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 111.8653618739806 ms 111.48762767348046 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6366.947154999991 ms 6352.16404099998 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 110.75350079333552 ms 110.16861786666747 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.64718350002386 ms 68.68160549998947 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.23221540559708 ms 36.22538058499631 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.90781724559397 ms 36.91675834190811 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5088.643873999899 ms 5082.261738499255 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.52543789987492 ms 88.498999799934 ms 1.02
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 62.63505049992091 ms 60.69171000035567 ms 1.03
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.241896824997205 ms 36.21554315710666 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30.85383222970745 ms 30.829622089502823 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15183.4274644998 ms 14952.524137000182 ms 1.02
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 214.62474969000746 ms 213.97269461333295 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 171.34324200014817 ms 170.41449899988947 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.42424965703262 ms 121.74922727076351 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.86750325174465 ms 127.51590181598549 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2419.94309050051 ms 2412.483122499907 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.9347341346632 ms 116.41906162533633 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.5074859999404 ms 83.04159249973964 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 18.040118916115937 ms 17.954717798726882 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.18995890872091 ms 16.191531535676557 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15947.305339000195 ms 16020.202677499583 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 284.8694550293367 ms 253.06591161863494 ms 1.13
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 162.6065769996785 ms 155.56535249925219 ms 1.05
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 147.04902205617466 ms 146.14549104973028 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 135.04960825590788 ms 134.71175487648534 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6065.240653999922 ms 5996.506612500014 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 127.67534472666966 ms 127.50568606667495 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 88.65249250004581 ms 88.20934350001153 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.47487586912959 ms 40.21830872028567 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.27212183742507 ms 40.20192526693389 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1928.7831195001672 ms 1928.175570500116 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.55902658666128 ms 95.48966796332327 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 58.22271750048458 ms 58.23218350042225 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.32241293317558 ms 13.258276658293926 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.605839049654502 ms 12.51521060171141 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59655.89287549994 ms 60430.702900000026 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37568.667386154004 ms 38304.701958351994 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36846.219718499924 ms 37548.55232750003 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 97.87371614247901 ms 97.69063523534015 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.8946553446159 ms 100.23740548629765 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6079.166581499976 ms 6053.155197000024 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.0432068900036 ms 121.90334264000285 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.18983099987054 ms 81.16031500003373 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.674222356916054 ms 39.63900264773282 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.268007800389874 ms 39.268328228789144 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6112.7042135001375 ms 6098.667970999941 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 103.82918512667251 ms 104.38908700000866 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.30783099990367 ms 67.73955900007422 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34.604973556372265 ms 34.51027741501625 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.1040377102908 ms 35.03588377959485 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7775.121030000037 ms 7699.20140399995 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 156.30899060266498 ms 155.60859181600426 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 134.1737774999956 ms 126.44680549999521 ms 1.06
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.45028963865962 ms 57.35804097393765 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.13468994628042 ms 57.06640501978451 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1944.136977999733 ms 1940.8526524998706 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.47276826670834 ms 79.21267474664396 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.72888500038607 ms 39.89376600020478 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.120926211302567 ms 11.089562012528129 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.174795451768055 ms 11.173363486138213 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1833.2886819998748 ms 1828.8207245000194 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.54628826002772 ms 92.21822712335475 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56.28914200042345 ms 56.458014999861916 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.607432032428102 ms 12.639844913203472 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.868208461515167 ms 11.911983000670325 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 255207.64880749994 ms 254818.33761899997 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 240862.95858059733 ms 241011.8444389933 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 242202.8763794999 ms 242041.27989450013 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70.2477979698134 ms 68.88961185379736 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.26012851833015 ms 66.35204241527413 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 76245.76136950009 ms 75939.70493999995 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56817.59918345466 ms 56341.681981972004 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69838.31376050011 ms 69316.22169849992 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.67750043316943 ms 66.55430329170304 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.64907742351473 ms 64.33734177531761 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 61966.44442949946 ms 63194.19884299987 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26686.82494355602 ms 27375.426378377993 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24798.986848500135 ms 25736.021327001254 ms 0.96
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 185.44655955972675 ms 185.86486330113644 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 197.25141847921026 ms 197.56442768654986 ms 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'smaller_is_better'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.10.

Benchmark suite Current: 3a25456 Previous: d485d3e Ratio
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 284.8694550293367 ms 253.06591161863494 ms 1.13

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: 3a25456 Previous: d485d3e Ratio
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.660780166138813 prompts/s 5.651181668821501 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2646.754374479863 tokens/s 2642.2665010741807 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7608517222511803 prompts/s 0.7581851684336032 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 98.91072389265344 tokens/s 98.56407189636842 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.497863240110499 prompts/s 7.491801833614156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3846.403842176686 tokens/s 3843.2943406440622 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.263165670937374 prompts/s 7.2590351137115885 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3395.9657411034786 tokens/s 3394.0344577669903 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.205290245587013 prompts/s 3.2034136753545686 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1510.864841901368 tokens/s 1509.9802904971816 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0091416547967595 prompts/s 2.0090910889604428 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4116.731250678561 tokens/s 4116.627641279947 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.921212009499216 prompts/s 13.935092429466788 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3577.7514864412988 tokens/s 3581.3187543729646 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 28.60050956981981 prompts/s 28.675999289815444 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3689.4657345067553 tokens/s 3699.2039083861923 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0414741186437655 prompts/s 2.041595395165156 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4182.980469101076 tokens/s 4183.228964693405 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8003552492032213 prompts/s 1.8008286607629982 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.04618239641877 tokens/s 234.10772589918977 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.46613907478029054 prompts/s 0.46622566694755396 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.0502210082543 tokens/s 126.07363668484456 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.11653395549375 tokens/s 118.1384758333972 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.228695962633237 prompts/s 2.228676699666547 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 688.2807451534806 tokens/s 688.2747962356874 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 511.60458720900175 tokens/s 511.6061084686538 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.292196273985361 prompts/s 10.296198262914897 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1337.985515618097 tokens/s 1338.5057741789367 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.983799175194995 prompts/s 0.9838550931346398 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.0764661368283 tokens/s 290.09295372772567 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.34321397408775 tokens/s 212.35528330218065 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8298168160888668 prompts/s 1.8273191772228683 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 581.388916853793 tokens/s 580.5953403918068 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 409.71550316833554 tokens/s 409.15625518475724 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14.277102337452854 prompts/s 14.267524211368551 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3669.2153007253837 tokens/s 3666.7537223217178 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.49420327447342 prompts/s 24.443938891144068 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3159.7522224070713 tokens/s 3153.268116957585 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.765123749090927 prompts/s 3.7646441882254944 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.2518428182 tokens/s 3858.760292931132 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.7252435380497463 prompts/s 0.7234133412978745 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 94.28165994646702 tokens/s 94.04373436872369 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9475074250197076 prompts/s 0.947734868686396 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 279.3757226251442 tokens/s 279.44278514841284 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 204.52895276475408 tokens/s 204.5812078708744 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.05806622022513 prompts/s 4.0550726389644725 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4159.5178757307585 tokens/s 4156.4494549385845 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.795442448447371 prompts/s 2.613947703296607 prompts/s 0.94
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 363.40751829815827 tokens/s 339.81320142855895 tokens/s 0.94
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.493490348755937 prompts/s 7.4912303349518945 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3844.160548911795 tokens/s 3843.001161830322 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.399469903448129 prompts/s 3.399009142352374 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.93108744825673 tokens/s 441.8711885058086 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.765087727808706 prompts/s 3.764889148332317 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.214921003924 tokens/s 3859.011377040625 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7287331590671284 prompts/s 3.7273345040302686 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 484.7353106787267 tokens/s 484.55348552393497 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.475168183286417 prompts/s 17.45885380130362 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2271.7718638272345 tokens/s 2269.6509941694703 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9681342060283681 prompts/s 0.9681604070981638 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 288.6750304395187 tokens/s 288.68284298717344 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 210.05930579599513 tokens/s 210.06499072944558 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.119431708948666 prompts/s 8.11569046610624 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4165.268466690665 tokens/s 4163.349209112502 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 9.906723419012536 prompts/s 9.893361252236701 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1287.8740444716298 tokens/s 1286.1369627907711 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 48.02672591354476 prompts/s 48.146230502846066 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3121.7371843804094 tokens/s 3129.5049826849945 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0420685187863388 prompts/s 2.0417904814497367 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4184.198394993208 tokens/s 4183.628696490511 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.080983750590903 prompts/s 26.027830568603793 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1695.2639437884088 tokens/s 1691.8089869592466 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.57543775801962 prompts/s 26.106528472473716 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1662.4034542712755 tokens/s 1696.9243507107915 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.8039408457872004 prompts/s 1.8018837875142801 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.51230995233607 tokens/s 234.24489237685643 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8717978054146744 prompts/s 3.872757050662234 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3968.5927505500413 tokens/s 3969.5759769287897 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919090298433429 prompts/s 0.4918963276896557 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03123294878927 tokens/s 130.0278752614836 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.85919929367331 tokens/s 121.99356857588587 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 8.034137241969928 prompts/s 8.040437806110551 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4121.512405130573 tokens/s 4124.7445945347135 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4750588150092445 prompts/s 0.475098682045607 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 125.5770471595437 tokens/s 125.58758561193576 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 99.44564527526852 tokens/s 99.4539907748804 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 49.515321747440886 prompts/s 49.29705602884645 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3218.4959135836575 tokens/s 3204.308641875019 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.905686185013021 prompts/s 13.969617172337554 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3573.761349548346 tokens/s 3590.1916132907513 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.9275308703473772 prompts/s 2.9263246157622866 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 380.579013145159 tokens/s 380.4222000490973 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9142154137323234 prompts/s 0.9137459158052325 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.84800378520205 tokens/s 118.78696905468021 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.801358920265668 prompts/s 1.8026610572464896 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 234.17665963453683 tokens/s 234.34593744204363 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.4822451989755328 prompts/s 3.4777941152484564 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1103.4886811033566 tokens/s 1102.0781771810832 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 795.9832150658239 tokens/s 794.9680900399532 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.451744162134273 prompts/s 2.4517680963310378 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 757.1639771113871 tokens/s 757.1713686295933 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 562.9368045871099 tokens/s 562.997873448432 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.6950954475147046 prompts/s 1.7036629049511658 prompts/s 1.01
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 552.833908997919 tokens/s 555.6280767196934 tokens/s 1.01
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 385.27937432924887 tokens/s 387.23576364378016 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.24076566594390642 prompts/s 0.2408087883142535 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.299536572707837 tokens/s 31.305142480852957 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.985618040953601 prompts/s 1.985354238826718 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 258.13034532396813 tokens/s 258.09605104747334 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.354068852708809 prompts/s 2.3557112612966504 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 726.9992368858859 tokens/s 727.5064564553736 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 540.2682179720825 tokens/s 540.6137478291425 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.962218323378384 prompts/s 15.931149649188745 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4102.290109108244 tokens/s 4094.3054598415074 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.886525889505889 prompts/s 7.891096679323421 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4045.787781316521 tokens/s 4048.1325964929147 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26.09702021354244 prompts/s 26.034278082038465 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1696.3063138802586 tokens/s 1692.2280753325003 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.3988450838261195 prompts/s 3.403168528141097 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.8498608973955 tokens/s 442.4119086583426 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839455562815028 prompts/s 0.983977399518168 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.11962708812206 tokens/s 290.1290161725969 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.38792814188332 tokens/s 212.2144057540833 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15.584781522763677 prompts/s 15.578473913603274 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4005.288851350265 tokens/s 4003.667795796041 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.494232195332185 prompts/s 7.486450852726426 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3844.541116205411 tokens/s 3840.5492874486563 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.179605846199559 prompts/s 4.17446080449878 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1324.4752966021783 tokens/s 1322.8448843376184 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 834.5084624638964 tokens/s 833.4895420694445 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 29.166654151072496 prompts/s 29.174460501148605 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3762.498385488352 tokens/s 3763.50540464817 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.658096038753769 prompts/s 5.653633475977627 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2645.4993838797122 tokens/s 2643.4128680280996 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.599437245474968 prompts/s 11.605228976145973 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1507.926841911746 tokens/s 1508.6797668989766 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.741491027541746 prompts/s 6.735574667216951 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 876.393833580427 tokens/s 875.6247067382036 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.041897561290103 prompts/s 2.0420892274236246 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4183.848103083421 tokens/s 4184.2408269910065 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.852266843018082 prompts/s 16.82528216827471 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2190.7946895923506 tokens/s 2187.2866818757125 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.380807074947154 prompts/s 5.374808985672395 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 699.50491974313 tokens/s 698.7251681374114 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.24059807885840614 prompts/s 0.24059288698853612 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.2777502515928 tokens/s 31.277075308509698 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 25.493687164919958 prompts/s 25.49805802814238 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1657.0896657197973 tokens/s 1657.3737718292548 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.566260481090202 prompts/s 6.5709542389398585 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 853.6138625417263 tokens/s 854.2240510621816 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.8373039483168765 prompts/s 3.8408865302828166 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1473.5247161536804 tokens/s 1474.9004276286016 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.579593866862496 prompts/s 11.588339878756337 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1505.3472026921243 tokens/s 1506.4841842383237 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.70729684767236 prompts/s 20.702290603550058 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2691.9485901974067 tokens/s 2691.2977784615077 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.7650914135293774 prompts/s 3.7626156458153983 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3859.2186988676117 tokens/s 3856.6810369607833 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.391485471986067 prompts/s 4.376621131542948 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1391.6178312176648 tokens/s 1386.9074703746448 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1003.0152818016178 tokens/s 999.9733138823539 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.4522833855895545 prompts/s 4.44196346527194 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1410.884082059474 tokens/s 1407.6138025100254 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1017.5426540761792 tokens/s 1015.4624612509341 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.53243208927685 prompts/s 3.5137946523034143 prompts/s 0.99
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1119.392404770941 tokens/s 1113.486387368429 tokens/s 0.99
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 807.5351702012235 tokens/s 803.2862529333154 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.749580479085355 prompts/s 6.748043912141075 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 877.4454622810962 tokens/s 877.2457085783398 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.956786133549555 prompts/s 13.959139226272045 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3586.8940363222355 tokens/s 3587.4987811519154 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.180220989305383 prompts/s 24.186631400191875 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3119.2485076203943 tokens/s 3120.0754506247517 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1.0157730895986503 prompts/s 1.0166856525214367 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.05050164782455 tokens/s 132.16913482778676 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9145609745625481 prompts/s 0.9143859155159352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.89292669313126 tokens/s 118.87016901707158 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 10.404749381039107 prompts/s 10.366746516695056 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1352.6174195350839 tokens/s 1347.6770471703574 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0643363682231834 prompts/s 2.0644265494330707 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4229.825218489303 tokens/s 4230.009999788362 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9149217262537348 prompts/s 0.9141892722282193 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.93982441298552 tokens/s 118.84460538966852 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.300755513543784 prompts/s 2.300974144441723 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 710.5346560626817 tokens/s 710.6021751141226 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 460.17871177491935 tokens/s 460.6703635448626 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 20.257196925084852 prompts/s 20.247217206011417 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2633.4356002610307 tokens/s 2632.138236781484 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.2567616430140824 prompts/s 0.25672487636671604 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 33.37901359183071 tokens/s 33.37423392767309 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3.39708790096349 prompts/s 3.3993142849262794 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 441.6214271252537 tokens/s 441.9108570404164 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.0728298553365345 prompts/s 2.0726681559409985 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4247.228373584559 tokens/s 4246.897051523106 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.9839026397951075 prompts/s 0.9839203228160437 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 290.10697301905344 tokens/s 290.11218691671985 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 212.37538479977394 tokens/s 212.38904088307117 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4.174023940387348 prompts/s 4.17260055673134 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 4278.374538897031 tokens/s 4276.915570649624 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.480771408304268 prompts/s 24.43199699436081 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3158.0195116712507 tokens/s 3151.7276122725443 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6.450842999015345 prompts/s 6.457116453290042 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 838.6095898719948 tokens/s 839.4251389277055 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4640366162447238 prompts/s 0.463520024094734 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.6634391381303 tokens/s 122.52688316920198 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.09345513845804 tokens/s 114.96532650946989 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.4919091812565093 prompts/s 0.4919132458930308 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.03127297334566 tokens/s 130.03234741936376 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.99019755707259 tokens/s 121.99448498147164 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.6617668429125025 prompts/s 5.656258130386432 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2647.21570507217 tokens/s 2644.64005144348 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.571682431902945 prompts/s 11.583829296115017 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1504.3187161473827 tokens/s 1505.8978084949522 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5.613483083398872 prompts/s 5.624589197137983 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 729.7528008418534 tokens/s 731.1965956279378 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.769269983663158 prompts/s 24.499635371471232 prompts/s 1.03
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3066.2358278925476 tokens/s 3160.4529629197887 tokens/s 1.03
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7.05701649797792 prompts/s 7.042005918148534 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3299.5786337945565 tokens/s 3292.560287089528 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2.447156473776068 prompts/s 2.4469783979387754 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 755.7471766080171 tokens/s 755.6921820407722 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 561.4331639643023 tokens/s 560.9322773920293 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 0.24058543627008597 prompts/s 0.2406090203706557 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 31.276106715111176 tokens/s 31.27917264818524 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: 3a25456 Previous: d485d3e Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6101.23209849985 ms 6098.667970999941 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 103.81138589331081 ms 104.38908700000866 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.00526649986205 ms 67.73955900007422 ms 0.94
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 34.55243954826612 ms 34.51027741501625 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 35.05923950958349 ms 35.03588377959485 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 7680.819542500103 ms 7699.20140399995 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 155.6045172239992 ms 155.60859181600426 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 130.96245799999906 ms 126.44680549999521 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57.10173666541715 ms 57.35804097393765 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56.79671118557445 ms 57.06640501978451 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 3649.93440049966 ms 3631.2452459997075 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 142.4122476466558 ms 143.76172412334503 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.2783384999093 ms 95.35134399993694 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24.36573514002923 ms 24.310515078398957 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 23.69197726093354 ms 23.651293949528196 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 254531.2522874997 ms 254818.33761899997 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 240153.88880120867 ms 241011.8444389933 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 241515.4948259999 ms 242041.27989450013 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 69.14310567244951 ms 68.88961185379736 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 66.18372669169628 ms 66.35204241527413 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6043.307231999961 ms 5996.506612500014 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 126.88578150666443 ms 127.50568606667495 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 82.40760800003955 ms 88.20934350001153 ms 0.93
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.35962738190891 ms 40.21830872028567 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 40.23074240547815 ms 40.20192526693389 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6041.992879999725 ms 6053.155197000024 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 121.46525663999304 ms 121.90334264000285 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.24388900003032 ms 81.16031500003373 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.55438731048751 ms 39.63900264773282 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.16477230530249 ms 39.268328228789144 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1942.0991919996595 ms 1940.8526524998706 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 78.98660993330243 ms 79.21267474664396 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37.880631999996695 ms 39.89376600020478 ms 0.95
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.054870860215237 ms 11.089562012528129 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.121548043603779 ms 11.173363486138213 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5049.914612500288 ms 5082.261738499255 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 90.22882873998849 ms 88.498999799934 ms 1.02
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 62.91983600021922 ms 60.69171000035567 ms 1.04
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.16473733373438 ms 36.21554315710666 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 30.799053717961876 ms 30.829622089502823 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 59942.670705500066 ms 60430.702900000026 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 38040.75315246733 ms 38304.701958351994 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 37545.920685000056 ms 37548.55232750003 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 97.56689999797372 ms 97.69063523534015 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 100.25388810633912 ms 100.23740548629765 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2411.809580999943 ms 2412.483122499907 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 115.5320580693345 ms 116.41906162533633 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 79.90481249953518 ms 83.04159249973964 ms 0.96
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.94385816744697 ms 17.954717798726882 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 16.169444361253195 ms 16.191531535676557 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 76860.2623894999 ms 75939.70493999995 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 57203.97092948268 ms 56341.681981972004 ms 1.02
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 70173.28971250003 ms 69316.22169849992 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 67.93434489798318 ms 66.55430329170304 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 64.58440380578197 ms 64.33734177531761 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5330.106188499485 ms 5320.154192500013 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 182.17156377735228 ms 181.2350845293695 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 172.8889075002371 ms 171.38535149933887 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.881822445030636 ms 39.67783502514967 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.93948749345439 ms 36.79312972752162 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1920.0154015002227 ms 1928.175570500116 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 95.74568158999682 ms 95.48966796332327 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 60.970186999838916 ms 58.23218350042225 ms 1.05
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13.202795011221378 ms 13.258276658293926 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.448855716742237 ms 12.51521060171141 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 15703.07831700029 ms 16020.202677499583 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 224.3258602553542 ms 253.06591161863494 ms 0.89
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 150.11075299935328 ms 155.56535249925219 ms 0.96
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 144.76176657570113 ms 146.14549104973028 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 132.821358348712 ms 134.71175487648534 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 14704.967883500103 ms 14952.524137000182 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 215.22999095600426 ms 213.97269461333295 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 167.70195599974613 ms 170.41449899988947 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 117.65424397527357 ms 121.74922727076351 ms 0.97
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 122.98717967737052 ms 127.51590181598549 ms 0.96
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 13476.068847999613 ms 13589.161211999908 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 199.25043000799937 ms 203.4137675746612 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 154.84395100020265 ms 159.45165350012758 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 105.13777527980294 ms 106.80662248274147 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 109.68111962814491 ms 111.48762767348046 ms 0.98
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 61124.359150999226 ms 63194.19884299987 ms 0.97
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 26456.67674058268 ms 27375.426378377993 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 24940.77680600003 ms 25736.021327001254 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 185.534421332363 ms 185.86486330113644 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 197.434688678055 ms 197.56442768654986 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 5879.7700890008855 ms 5856.753418500375 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 118.73317806397729 ms 118.7610248973045 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 87.43678700011515 ms 89.27816299910774 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 53.14737516562541 ms 53.309391464699814 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 46.20036354770608 ms 46.510393959855044 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 1818.5470739995253 ms 1828.8207245000194 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 92.5462979966657 ms 92.21822712335475 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 56.357280499923945 ms 56.458014999861916 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 12.59746519769587 ms 12.639844913203472 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.822087371719496 ms 11.911983000670325 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 6362.97378599994 ms 6352.16404099998 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 109.44010315333193 ms 110.16861786666747 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 68.66087950004385 ms 68.68160549998947 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.251718120805435 ms 36.22538058499631 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 36.98496586314468 ms 36.91675834190811 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2032.796726000015 ms 2033.819562499957 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 80.79107167337497 ms 82.19037280665968 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 39.870010999948136 ms 38.89878250038237 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.590794727441683 ms 11.560121123356186 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 11.656013017847151 ms 11.691221148500436 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 2563.39438949999 ms 2571.6090910000275 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 119.95549527066639 ms 120.15518762534703 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 83.83280099997137 ms 81.6943180002454 ms 1.03
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 19.130742047195096 ms 19.216760619551575 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, May 10 2024, 13:42:25) [GCC 9.4.0]", "torch_version": "2.3.0+cu121"} 17.251971973906564 ms 17.232633518539803 ms 1.00

This comment was automatically generated by workflow using github-action-benchmark.

Please sign in to comment.