Bump to v2.3.2

README.md

@@ -31,7 +31,7 @@ Please cite and credit FlashAttention if you use it.
 Requirements:
 - CUDA 11.6 and above.
 - PyTorch 1.12 and above.
-- Linux. Windows is not supported for now. If you have ideas on how to modify the code to support Windows, please reach out via Github issue.
+- Linux. Might work for Windows starting v2.3.2 (we've seen a few positive [reports](https://github.com/Dao-AILab/flash-attention/issues/595)) but Windows compilation still requires more testing. If you have ideas on how to set up prebuilt CUDA wheels for Windows, please reach out via Github issue.
 
 We recommend the
 [Pytorch](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch)

flash_attn/__init__.py

-__version__ = "2.3.1.post1"
+__version__ = "2.3.2"
 
 from flash_attn.flash_attn_interface import (
     flash_attn_func,

training/Dockerfile

@@ -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==2.3.1.post1
+RUN pip install flash-attn==2.3.2
 
 # Install CUDA extensions for fused dense, layer norm
 RUN git clone https://github.com/HazyResearch/flash-attention \
-    && cd flash-attention && git checkout v2.3.1.post1 \
+    && cd flash-attention && git checkout v2.3.2 \
     && cd csrc/layer_norm && pip install . && cd ../../ \
     && cd csrc/fused_dense_lib && pip install . && cd ../../ \
     && cd .. && rm -rf flash-attention