Try switching back to Cutlass 3.2.0

cutlass @ 3a8f57a3

-Subproject commit 34fd98056b69fbf7f0929b3f734bb5f00642e2c9
+Subproject commit 3a8f57a3c89cfff7aa686e95f13d9ad850f61898

flash_attn/__init__.py

-__version__ = "2.1.2"
+__version__ = "2.1.2.post1"
 
 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==2.1.2
+RUN pip install flash-attn==2.1.2.post1
 
 # Install CUDA extensions for cross-entropy, fused dense, layer norm
 RUN git clone https://github.com/HazyResearch/flash-attention \
-    && cd flash-attention && git checkout v2.1.2 \
+    && cd flash-attention && git checkout v2.1.2.post1 \
     && cd csrc/fused_softmax && pip install . && cd ../../ \
     && cd csrc/rotary && pip install . && cd ../../ \
     && cd csrc/xentropy && pip install . && cd ../../ \