Skip to content

GitLab

Commit 1d57ec3d authored by zhuwenwen's avatar zhuwenwen
Browse files

update vdim pad

parent 4564b2f5
Show whitespace changes
Inline Side-by-side
Showing with 1 addition and 1 deletion
vllm/attention/backends/mla/utils.py
View file @ 1d57ec3d
...@@ -535,7 +535,7 @@ class MLACommonImpl(MLAAttentionImpl[T], Generic[T]): ...@@ -535,7 +535,7 @@ class MLACommonImpl(MLAAttentionImpl[T], Generic[T]):
# v with 0s to match the qk head dim # v with 0s to match the qk head dim
# v_padded = torch.nn.functional.pad(v, [0, q.shape[-1] - v.shape[-1]], # v_padded = torch.nn.functional.pad(v, [0, q.shape[-1] - v.shape[-1]],
# value=0) # value=0)
v_padded = torch.nn.functional.pad(v, [0, (q.shape[-1] - v.shape[-1]-32)], v_padded = torch.nn.functional.pad(v, [0, q.shape[-1] - v.shape[-1]],
value=0) value=0)
if torch.cuda.get_device_properties(torch.cuda.current_device()).multi_processor_count == 120: if torch.cuda.get_device_properties(torch.cuda.current_device()).multi_processor_count == 120:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment