biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Attention][2/n] Remove usage of deprecated seq_lens_cpu and num_computed_tokens_cpu CommonAttentionMetadata properties (#31774)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This commit is contained in:
Lucas Wilkinson
2026-01-06 12:32:14 -05:00
committed by GitHub
parent 2f4bdee61e
commit 4c73be14e0
3 changed files with 13 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/v1/attention/backends/gdn_attn.py
View File

@@ -142,8 +142,7 @@ class GDNAttentionMetadataBuilder(AttentionMetadataBuilder[GDNAttentionMetadata]
m = common_attn_metadata
query_start_loc = m.query_start_loc
context_lens = m.num_computed_tokens_cpu
context_lens_tensor = context_lens.to(query_start_loc.device, non_blocking=True)
context_lens_tensor = m.compute_num_computed_tokens()
nums_dict, batch_ptr, token_chunk_offset_ptr = None, None, None
if (
@@ -370,6 +369,5 @@ class GDNAttentionMetadataBuilder(AttentionMetadataBuilder[GDNAttentionMetadata]
num_accepted_tokens = torch.diff(m.query_start_loc)
num_decode_draft_tokens_cpu = (num_accepted_tokens - 1).cpu()
m._num_computed_tokens_cpu = m.seq_lens_cpu - num_accepted_tokens.cpu()
return self.build(0, m, num_accepted_tokens, num_decode_draft_tokens_cpu)