Overlap gradient computation and NCCL AllReduce #361
+94
−32
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
PeterZhizhin
force-pushed
the
add_nvcc_parallel
branch
2 times, most recently
from
0ffa9a0
to
61a1f15
Compare
PeterZhizhin
force-pushed
the
add_nvcc_parallel
branch
from
61a1f15
to
47cfb4c
Compare
ngc92
reviewed
May 17, 2024
| @@ -2348,7 +2410,7 @@ void common_free(GPT2 &model) { | |||
| cudaCheck(cudaFree(cublaslt_workspace)); | |||
| cublasCheck(cublasDestroy(cublas_handle)); | |||
| cublasCheck(cublasLtDestroy(cublaslt_handle)); | |||
| create_cudnn(); | |||
| destroy_cudnn(); | |||
There was a problem hiding this comment.
@karpathy @PeterZhizhin cherry pick; this should be merged immediately
ngc92
reviewed
May 17, 2024
| printf0("step %4d/%d: train loss %f (acc %f) (%f ms, %0f tok/s)\n", | ||
| step + 1, train_num_batches, model.mean_loss, accumulated_loss, | ||
| time_elapsed_ms, bias_corrected_ema_tokens_per_second); | ||
| logger_log_train(&logger, step, model.mean_loss); | ||
|
|
||
| // disable the profiler after 3 steps of optimization | ||
| if (step == 3) { cudaProfilerStop(); } | ||
| if (step == 3) { cudaCheck(cudaProfilerStop()); } |
There was a problem hiding this comment.
this is an independent fix too
ngc92
reviewed
May 22, 2024
| // Aggregate grads.lnfw and grads.lnfb in a background stream | ||
| floatX* layernorm_backward_pointers[] = {grads.lnfw, grads.lnfb}; | ||
| size_t layernorm_backward_sizes[] = {C, C}; | ||
| multi_gpu_async_all_reduce_pointers_group(2, layernorm_backward_pointers, layernorm_backward_sizes, multi_gpu_config, main_stream); |
There was a problem hiding this comment.
comment says background stream, but call uses main_stream?
There was a problem hiding this comment.
oh wait, in the version this code was based on, main_stream was the background stream?
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.
On my setup, I get the following:
Before:
After:
So, a 12% speedup.
NSight Systems profiles:
Before:
After:
