From e12bc85014ae0a24f8a0b9f84c2cbde985c872ff Mon Sep 17 00:00:00 2001 From: chenyu Date: Mon, 18 Mar 2024 23:49:55 -0400 Subject: [PATCH] use BS=128 and BS=768 for resent benchmark (#3815) 50% more hcopt perf with this one weird trick --- .github/workflows/benchmark.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/benchmark.yml b/.github/workflows/benchmark.yml index 6a379e55dd..ff14263b2b 100644 --- a/.github/workflows/benchmark.yml +++ b/.github/workflows/benchmark.yml @@ -222,9 +222,9 @@ jobs: - name: Run MLPerf resnet eval on training data run: time HSA=1 MODEL=resnet python3 examples/mlperf/model_eval.py - name: Run 10 MLPerf ResNet50 training steps (1 gpu) - run: HSA=1 BENCHMARK=10 BS=104 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py | tee train_resnet_one_gpu.txt + run: HSA=1 BENCHMARK=10 BS=128 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py | tee train_resnet_one_gpu.txt - name: Run 10 MLPerf ResNet50 training steps (6 gpu) - run: HSA=1 BENCHMARK=10 BS=624 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py | tee train_resnet.txt + run: HSA=1 BENCHMARK=10 BS=768 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py | tee train_resnet.txt - uses: actions/upload-artifact@v4 with: name: Speed (AMD Training)