From f7d5f7ff5f4fc70556e92204bb93bf3855b76948 Mon Sep 17 00:00:00 2001 From: kouta Date: Thu, 8 Aug 2024 00:29:57 +0900 Subject: [PATCH] fix fused attn error --- .../v3-megatron-sakura/scripts/environment.sh | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/pretrain/installers/v3-megatron-sakura/scripts/environment.sh b/pretrain/installers/v3-megatron-sakura/scripts/environment.sh index 25b0a2a..b119436 100644 --- a/pretrain/installers/v3-megatron-sakura/scripts/environment.sh +++ b/pretrain/installers/v3-megatron-sakura/scripts/environment.sh @@ -3,11 +3,12 @@ export PRETRAIN_CUDA_VERSION_MAJOR=12 export PRETRAIN_CUDA_VERSION_MINOR=1 -export PRETRAIN_CUDA_VERSION=${PRETRAIN_CUDA_VERSION_MAJOR}.${PRETRAIN_CUDA_VERSION_MINOR} +export PRETRAIN_CUDA_VERSION_PATCH=1 +export PRETRAIN_CUDA_VERSION=${PRETRAIN_CUDA_VERSION_MAJOR}.${PRETRAIN_CUDA_VERSION_MINOR}.${PRETRAIN_CUDA_VERSION_PATCH} export PRETRAIN_CUDA_VERSION_SHORT=${PRETRAIN_CUDA_VERSION_MAJOR}${PRETRAIN_CUDA_VERSION_MINOR} -export PRETRAIN_CUDNN_VERSION=8.9.4 +export PRETRAIN_CUDNN_VERSION=8.9.7 export PRETRAIN_HPCX_VERSION=2.17.1 -export PRETRAIN_NCCL_VERSION=2.20.5 +export PRETRAIN_NCCL_VERSION=2.18.3 export PRETRAIN_PYTHON_VERSION=3.10.14 export PRETRAIN_TORCH_VERSION=2.3.1 @@ -17,7 +18,8 @@ export PRETRAIN_TRANSFORMER_ENGINE_VERSION=1.4 export PRETRAIN_MEGATRON_TAG=nii-geniac export PRETRAIN_TOKENIZER_TAG=Release-ver3.0b1 -module load cuda/${PRETRAIN_CUDA_VERSION} -module load /data/cudnn-tmp-install/modulefiles/${PRETRAIN_CUDNN_VERSION} -module load hpcx/${PRETRAIN_HPCX_VERSION}-gcc-cuda${PRETRAIN_CUDA_VERSION_MAJOR}/hpcx -module load nccl/${PRETRAIN_NCCL_VERSION} +export MODULEPATH=/data/modules:${MODULEPATH} +module load cuda-${PRETRAIN_CUDA_VERSION} +module load cudnn-${PRETRAIN_CUDNN_VERSION} +module load hpcx-${PRETRAIN_HPCX_VERSION} +module load nccl-${PRETRAIN_NCCL_VERSION} \ No newline at end of file