Skip to content

GitLab

Unverified Commit 78237e43 authored by Michael Goin's avatar Michael Goin Committed by GitHub
Browse files

[Bugfix] Remove contiguous output req for context parallel MLA (#25414) 


Signed-off-by: default avatarMichael Goin <mgoin64@gmail.com>
parent eea17839
Show whitespace changes
Inline Side-by-side
Showing with 0 additions and 1 deletion
vllm/attention/ops/common.py
View file @ 78237e43
...@@ -134,6 +134,5 @@ def cp_lse_ag_out_rs(cp_attn_out: torch.Tensor, ...@@ -134,6 +134,5 @@ def cp_lse_ag_out_rs(cp_attn_out: torch.Tensor,
cp_attn_lse = cp_attn_lse.contiguous() cp_attn_lse = cp_attn_lse.contiguous()
lses = cp_group.all_gather(cp_attn_lse, dim=0).view_as(lses) lses = cp_group.all_gather(cp_attn_lse, dim=0).view_as(lses)
out, _ = correct_attn_out(cp_attn_out, lses, cp_group.rank_in_group, ctx) out, _ = correct_attn_out(cp_attn_out, lses, cp_group.rank_in_group, ctx)
assert out.is_contiguous()
out = cp_group.reduce_scatter(out, dim=1) out = cp_group.reduce_scatter(out, dim=1)
return out return out
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