disable autocast for ring_flash_attn_cuda

lucidrains · Apr 6, 2024 · 8225bfc · 8225bfc
1 parent 5b202e4
commit 8225bfc
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 1 deletion.
diff --git a/ring_attention_pytorch/ring_flash_attention_cuda.py b/ring_attention_pytorch/ring_flash_attention_cuda.py
@@ -7,6 +7,7 @@
 from torch import nn, einsum, Tensor
 import torch.nn.functional as F
 from torch.autograd.function import Function
+from torch.cuda.amp import autocast
 
 from ring_attention_pytorch.ring import (
     ring_pass,
@@ -833,6 +834,7 @@ def backward(ctx, do):
 
 ring_flash_attn_cuda_ = RingFlashAttentionCUDAFunction.apply
 
+@autocast(enabled = False)
 @beartype
 def ring_flash_attn_cuda(
     q: Tensor,

diff --git a/setup.py b/setup.py
@@ -3,7 +3,7 @@
 setup(
   name = 'ring-attention-pytorch',
   packages = find_packages(exclude=[]),
-  version = '0.3.0',
+  version = '0.3.2',
   license='MIT',
   description = 'Ring Attention - Pytorch',
   author = 'Phil Wang',