From a39adaaa6f1c8ac6a72ad0f4fcefaa25b5dbabe1 Mon Sep 17 00:00:00 2001 From: Chuck Tang Date: Thu, 14 Nov 2024 08:35:28 -0800 Subject: [PATCH 1/3] commit change --- docker/Dockerfile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docker/Dockerfile b/docker/Dockerfile index c3f4dee907..b20ddf1554 100644 --- a/docker/Dockerfile +++ b/docker/Dockerfile @@ -262,7 +262,7 @@ RUN if [ -n "$MOFED_VERSION" ] ; then \ RUN if [ -n "$CUDA_VERSION" ] ; then \ pip${PYTHON_VERSION} install --upgrade --no-cache-dir ninja==1.11.1 && \ pip${PYTHON_VERSION} install --upgrade --no-cache-dir --force-reinstall packaging==22.0 && \ - MAX_JOBS=1 pip${PYTHON_VERSION} install --no-cache-dir --no-build-isolation flash-attn==2.6.3; \ + pip${PYTHON_VERSION} install --no-cache-dir --no-build-isolation https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.0.post2/flash_attn-2.7.0.post2+cu12torch2.5cxx11abiTRUE-cp311-cp311-linux_x86_64.wh; \ cd .. ; \ fi From 060ef0bb1db5a56c33989aab8fb000b38dca10cc Mon Sep 17 00:00:00 2001 From: Chuck Tang Date: Thu, 14 Nov 2024 08:38:01 -0800 Subject: [PATCH 2/3] commit change --- docker/Dockerfile | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docker/Dockerfile b/docker/Dockerfile index b20ddf1554..6c549f5941 100644 --- a/docker/Dockerfile +++ b/docker/Dockerfile @@ -259,6 +259,8 @@ RUN if [ -n "$MOFED_VERSION" ] ; then \ ########################## # Install Flash Attention ########################## +# Use the right flash attention wheel for the current PyTorch and CUDA version +# https://github.com/Dao-AILab/flash-attention/releases RUN if [ -n "$CUDA_VERSION" ] ; then \ pip${PYTHON_VERSION} install --upgrade --no-cache-dir ninja==1.11.1 && \ pip${PYTHON_VERSION} install --upgrade --no-cache-dir --force-reinstall packaging==22.0 && \ From 83fac05b06efa2c5fb7eeed93c4b9e0562700b6b Mon Sep 17 00:00:00 2001 From: Chuck Tang Date: Thu, 14 Nov 2024 08:39:22 -0800 Subject: [PATCH 3/3] commit change --- docker/Dockerfile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docker/Dockerfile b/docker/Dockerfile index 6c549f5941..f34c5ad9e4 100644 --- a/docker/Dockerfile +++ b/docker/Dockerfile @@ -259,7 +259,7 @@ RUN if [ -n "$MOFED_VERSION" ] ; then \ ########################## # Install Flash Attention ########################## -# Use the right flash attention wheel for the current PyTorch and CUDA version +# Make sure to bump the flash attention wheel for the current PyTorch and CUDA version # https://github.com/Dao-AILab/flash-attention/releases RUN if [ -n "$CUDA_VERSION" ] ; then \ pip${PYTHON_VERSION} install --upgrade --no-cache-dir ninja==1.11.1 && \