Skip to content

fix multihead attention for half #21658

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
ngimel wants to merge 1 commit into from
Closed

fix multihead attention for half #21658

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions test/test_nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3495,6 +3495,21 @@ def test_multihead_attn_key_padding_mask():
test_multihead_attn_no_masking() # Test MultiheadAttention without masking
test_multihead_attn_key_padding_mask() # Test MultiheadAttention with src lengths

@repeat_test_for_types(ALL_TENSORTYPES)
@unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
def test_multihead_attention_dtype(self, dtype=torch.float):
embed_dim = 128
num_heads = 8
sl = 10
bs = 8
model = nn.MultiheadAttention(embed_dim, num_heads).cuda().to(dtype)
q = torch.randn(sl, bs, embed_dim, device="cuda", dtype=dtype)
k = torch.randn(sl, bs, embed_dim, device="cuda", dtype=dtype)
v = torch.randn(sl, bs, embed_dim, device="cuda", dtype=dtype)
out = model(q, k, v)
self.assertEqual(q.size(), out[0].size())
self.assertEqual(dtype, out[0].dtype)

def test_normalize(self):
inputs = torch.randn(1, 3, 4, 4, requires_grad=True)
self.assertTrue(gradcheck(lambda x: F.normalize(x, p=1, dim=-1), (inputs,)))
Expand Down
3 changes: 1 addition & 2 deletions torch/nn/functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3272,8 +3272,7 @@ def multi_head_attention_forward(query, # type: Tensor
attn_output_weights = attn_output_weights.view(bsz * num_heads, tgt_len, src_len)

attn_output_weights = softmax(
attn_output_weights.float(), dim=-1,
dtype=torch.float32 if attn_output_weights.dtype == torch.float16 else attn_output_weights.dtype)
attn_output_weights, dim=-1)
attn_output_weights = dropout(attn_output_weights, p=dropout_p, training=training)

attn_output = torch.bmm(attn_output_weights, v)
Expand Down