Skip to content
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

hotfix - Revert vllm/attention/layer.py changes from 0f8cafe - fix torch.compile recompilations #709

Closed
wants to merge 4 commits into from
Closed

hotfix - Revert vllm/attention/layer.py changes from 0f8cafe - fix torch.compile recompilations #709

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
daafd10
[SW-216341] Revert vllm/attention/layer.py changes from 0f8cafe - fix…
Jan 20, 2025
695c40f
Merge branch 'habana_main' into dev/rlitka/revert_0f8cafe_layer
RafLit Jan 20, 2025
64c3e6b
Merge branch 'habana_main' into dev/rlitka/revert_0f8cafe_layer
RafLit Jan 23, 2025
01eebd2
Merge branch 'habana_main' into dev/rlitka/revert_0f8cafe_layer
RafLit Jan 23, 2025
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
22 changes: 10 additions & 12 deletions vllm/attention/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,12 @@ def forward(
kv_cache: torch.Tensor,
attn_metadata: AttentionMetadata,
) -> torch.Tensor:
if self.use_output:

if self.use_direct_call:
return self.impl.forward(query, key, value, kv_cache,
attn_metadata, self._k_scale,
self._v_scale)
Comment on lines +139 to +142
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could we add only those lines and left the rest unattached?

elif self.use_output:
output = torch.empty_like(query)
hidden_size = query.size(-1)
# Reshape the query, key, and value tensors.
Expand All @@ -147,19 +152,12 @@ def forward(
key = key.view(-1, self.num_kv_heads, self.head_size)
if value is not None:
value = value.view(-1, self.num_kv_heads, self.head_size)
if self.use_direct_call:
unified_attention_with_output(query, key, value, output,
self.layer_name)
else:
torch.ops.vllm.unified_attention_with_output(
query, key, value, output, self.layer_name)
torch.ops.vllm.unified_attention_with_output(
query, key, value, output, self.layer_name)
return output.view(-1, hidden_size)
else:
if self.use_direct_call:
return unified_attention(query, key, value, self.layer_name)
else:
return torch.ops.vllm.unified_attention(
query, key, value, self.layer_name)
return torch.ops.vllm.unified_attention(query, key, value,
self.layer_name)

def extra_repr(self) -> str:
s = f"head_size={self.impl.head_size}" # type: ignore
Expand Down
Loading