Skip to content

GitLab

Unverified Commit 6c526794 authored by vcherepanov-nv's avatar vcherepanov-nv Committed by GitHub
Browse files

Bump up FA to 2.8.1 (#1949) 


Signed-off-by: default avatarVladimir Cherepanov <vcherepanov@nvidia.com>
parent e7251f93
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 3 deletions
qa/L3_pytorch_FA_versions_test/test.sh
View file @ 6c526794
...@@ -18,10 +18,10 @@ sm_arch=`python3 -c "import torch; sm = torch.cuda.get_device_capability(0); pri ...@@ -18,10 +18,10 @@ sm_arch=`python3 -c "import torch; sm = torch.cuda.get_device_capability(0); pri
export FLASH_ATTN_CUDA_ARCHS=$sm_arch export FLASH_ATTN_CUDA_ARCHS=$sm_arch
if [ $sm_arch -gt 90 ] if [ $sm_arch -gt 90 ]
then then
FA_versions=(2.7.3) FA_versions=(2.8.1)
elif [ $sm_arch -eq 90 ] elif [ $sm_arch -eq 90 ]
then then
FA_versions=(2.5.7 2.7.3 3.0.0b1) FA_versions=(2.7.3 2.8.1 3.0.0b1)
fi fi
for fa_version in "${FA_versions[@]}" for fa_version in "${FA_versions[@]}"
......
transformer_engine/pytorch/attention/dot_product_attention/utils.py
View file @ 6c526794
...@@ -105,7 +105,7 @@ class FlashAttentionUtils: ...@@ -105,7 +105,7 @@ class FlashAttentionUtils:
version = PkgVersion("0") version = PkgVersion("0")
version_required = PkgVersion("2.1.1") version_required = PkgVersion("2.1.1")
version_required_blackwell = PkgVersion("2.7.3") version_required_blackwell = PkgVersion("2.7.3")
max_version = PkgVersion("2.7.4.post1") max_version = PkgVersion("2.8.1")
v2_plus = False v2_plus = False
v2_1_plus = False v2_1_plus = False
v2_3_plus = False v2_3_plus = False
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment