squall/flash-attention
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Re-enable compilation for Hopper

Browse Source
This commit is contained in:
Tri Dao 2023-09-21 23:55:25 -07:00
parent 187c2a0635
commit bff3147175
3 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
flash_attn/__init__.py
View File

@ -1,4 +1,4 @@
__version__ = "2.2.4"
__version__ = "2.2.4.post1"
from flash_attn.flash_attn_interface import (
flash_attn_func,

8
setup.py
View File

@ -122,10 +122,10 @@ if not SKIP_CUDA_BUILD:
# cc_flag.append("arch=compute_75,code=sm_75")
cc_flag.append("-gencode")
cc_flag.append("arch=compute_80,code=sm_80")
# if CUDA_HOME is not None:
# if bare_metal_version >= Version("11.8"):
# cc_flag.append("-gencode")
# cc_flag.append("arch=compute_90,code=sm_90")
if CUDA_HOME is not None:
if bare_metal_version >= Version("11.8"):
cc_flag.append("-gencode")
cc_flag.append("arch=compute_90,code=sm_90")
# HACK: The compiler flag -D_GLIBCXX_USE_CXX11_ABI is set to be the same as
# torch._C._GLIBCXX_USE_CXX11_ABI

4
training/Dockerfile
View File

@ -85,11 +85,11 @@ RUN pip install transformers==4.25.1 datasets==2.8.0 pytorch-lightning==1.8.6 tr
RUN pip install git+https://github.com/mlcommons/logging.git@2.1.0
# Install FlashAttention
RUN pip install flash-attn==2.2.4
RUN pip install flash-attn==2.2.4.post1
# Install CUDA extensions for fused dense, layer norm
RUN git clone https://github.com/HazyResearch/flash-attention \
&& cd flash-attention && git checkout v2.2.4 \
&& cd flash-attention && git checkout v2.2.4.post1 \
&& cd csrc/layer_norm && pip install . && cd ../../ \
&& cd csrc/fused_dense_lib && pip install . && cd ../../ \
&& cd .. && rm -rf flash-attention