From def90b22f67a9d7ce697e894449f05d4e7f5ee91 Mon Sep 17 00:00:00 2001 From: chenyu Date: Sun, 12 Jan 2025 09:43:56 -0500 Subject: [PATCH] EVAL_BS=36 for bert [pr] (#8576) 3X faster eval compared to BS=6. green https://wandb.ai/chenyuxyz/MLPerf-BERT/runs/ka5p5sm9/overview red https://wandb.ai/chenyuxyz/MLPerf-BERT/runs/a7maxsxd/overview --- .../benchmarks/bert/implementations/tinybox_green/dev_beam.sh | 2 +- .../benchmarks/bert/implementations/tinybox_green/dev_run.sh | 2 +- .../bert/implementations/tinybox_green/run_and_time.sh | 2 +- .../benchmarks/bert/implementations/tinybox_red/dev_beam.sh | 2 +- .../benchmarks/bert/implementations/tinybox_red/dev_run.sh | 2 +- .../benchmarks/bert/implementations/tinybox_red/run_and_time.sh | 2 +- 6 files changed, 6 insertions(+), 6 deletions(-) diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_beam.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_beam.sh index a6b45747f7..7318b3bb29 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_beam.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_beam.sh @@ -2,7 +2,7 @@ export PYTHONPATH="." export MODEL="bert" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=4 BEAM_UOPS_MAX=2000 BEAM_UPCAST_MAX=64 BEAM_LOCAL_MAX=512 export IGNORE_JIT_FIRST_BEAM=1 diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_run.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_run.sh index a1bb814f07..555ed80cb4 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_run.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/dev_run.sh @@ -2,7 +2,7 @@ export PYTHONPATH="." export MODEL="bert" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=4 BEAM_UOPS_MAX=2000 BEAM_UPCAST_MAX=64 BEAM_LOCAL_MAX=512 export IGNORE_JIT_FIRST_BEAM=1 diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/run_and_time.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/run_and_time.sh index 50542450c9..4e732799ae 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/run_and_time.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/run_and_time.sh @@ -3,7 +3,7 @@ export PYTHONPATH="." export MODEL="bert" export SUBMISSION_PLATFORM="tinybox_green" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=4 BEAM_UOPS_MAX=2000 BEAM_UPCAST_MAX=64 BEAM_LOCAL_MAX=512 export IGNORE_JIT_FIRST_BEAM=1 diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_beam.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_beam.sh index 5af5fd2d5f..692e0b49a1 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_beam.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_beam.sh @@ -2,7 +2,7 @@ export PYTHONPATH="." export MODEL="bert" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=3 export IGNORE_JIT_FIRST_BEAM=1 diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_run.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_run.sh index a4636ac6f9..86221ca594 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_run.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/dev_run.sh @@ -2,7 +2,7 @@ export PYTHONPATH="." export MODEL="bert" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=3 export IGNORE_JIT_FIRST_BEAM=1 diff --git a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/run_and_time.sh b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/run_and_time.sh index 458fcac97a..272b0ddb12 100755 --- a/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/run_and_time.sh +++ b/examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/run_and_time.sh @@ -3,7 +3,7 @@ export PYTHONPATH="." export MODEL="bert" export SUBMISSION_PLATFORM="tinybox_red" -export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=6 +export DEFAULT_FLOAT="HALF" GPUS=6 BS=72 EVAL_BS=36 export BEAM=3 export IGNORE_JIT_FIRST_BEAM=1