From 16838eae080fddb6c8c4e00e9c448421c66d0e2e Mon Sep 17 00:00:00 2001 From: Francis Lam Date: Tue, 30 Apr 2024 15:08:12 -0700 Subject: [PATCH] mlperf/resnet: update tinybox_red parameters to new best values (#4364) about 27 minutes to setup and 345ms/110TF steps --- .../resnet/implementations/tinybox_red/run_and_time.sh | 2 +- .../benchmarks/resnet/implementations/tinybox_red/setup.sh | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/run_and_time.sh b/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/run_and_time.sh index 8e387bc91a..ddd7152b34 100755 --- a/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/run_and_time.sh +++ b/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/run_and_time.sh @@ -4,7 +4,7 @@ export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=48 LR=7 export SPLIT_REDUCEOP=1 LAZYCACHE=0 RESET_STEP=0 -export TRAIN_BEAM=2 IGNORE_JIT_FIRST_BEAM=1 BEAM_UOPS_MAX=1200 BEAM_UPCAST_MAX=64 BEAM_LOCAL_MAX=256 BEAM_MIN_PROGRESS=50 +export TRAIN_BEAM=4 IGNORE_JIT_FIRST_BEAM=1 BEAM_UOPS_MAX=1500 BEAM_UPCAST_MAX=128 BEAM_LOCAL_MAX=1024 BEAM_MIN_PROGRESS=25 export WANDB=1 PARALLEL=0 diff --git a/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/setup.sh b/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/setup.sh index 7a0f31d283..41197eb53e 100755 --- a/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/setup.sh +++ b/examples/mlperf/training_submission_v4.0/tinycorp/benchmarks/resnet/implementations/tinybox_red/setup.sh @@ -4,7 +4,7 @@ export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=48 LR=7 export SPLIT_REDUCEOP=1 LAZYCACHE=0 RESET_STEP=0 -export TRAIN_BEAM=2 IGNORE_JIT_FIRST_BEAM=1 BEAM_UOPS_MAX=1200 BEAM_UPCAST_MAX=64 BEAM_LOCAL_MAX=256 BEAM_MIN_PROGRESS=50 +export TRAIN_BEAM=4 IGNORE_JIT_FIRST_BEAM=1 BEAM_UOPS_MAX=1500 BEAM_UPCAST_MAX=128 BEAM_LOCAL_MAX=1024 BEAM_MIN_PROGRESS=25 export BENCHMARK=10 DEBUG=2