Add 'run_batch' mode for GPU encoding and decoding with batch_size >= 1 #1534
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ame() deprecated in 1.13.1
|
libtorch-gpu代码中，没有显式的释放显存。在调用量增加的时候，是否会存在out of memory的问题？ |
veelion
commented
Nov 21, 2022
| r_hyps_pad_sos_eos, ctc_scores_tensor).toTuple()->elements(); | ||
| auto rescores = outputs[1].toTensor().to(at::kCPU); | ||
| #ifdef USE_GPU | ||
| c10::cuda::CUDACachingAllocator::emptyCache(); |
There was a problem hiding this comment.
#1534 clear GPU memory cache here, so it could support much more concurrency.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This mode improves the throughput of websocket server.
Test result:
hardware-1:
Platinum 8358P CPU @ 2.60GHz 15 cores + 80G memory, A5000 * 1 + 24G memory
hardware-2:
Platinum 8369B CPU @ 2.90GHz 32 cores + 120GB memory, A100-SXM4-80GB * 1 + 80GB memory
data:
3000 wavs with different durations in range [0.6, 15] seconds.
With same CPU, GPU is 2~3 times faster than CPU, run_batch is 2.x times faster than non run_batch mode, but the CER has a little bigger.