Rain-Bus/nano-vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #213 from Anai-Guo/fix/prepare-prefill-seqlen-k-chunked-prefill

Browse Source 
fix(model_runner): correct seqlen_k to chunk boundary in prepare_prefill
This commit is contained in:
Xingkai Yu
2026-04-25 17:43:24 +08:00
committed by GitHub
parent 812eb1c1e4 25794a1f29
commit 5df84934f3
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nanovllm/engine/model_runner.py
+1 -1
View File
@@ -139,8 +139,8 @@ class ModelRunner:
seqlen = len(seq) seqlen = len(seq)
start = min(seq.num_cached_tokens, seqlen - 1) start = min(seq.num_cached_tokens, seqlen - 1)
seqlen_q = seq.num_scheduled_tokens seqlen_q = seq.num_scheduled_tokens
seqlen_k = seqlen
end = start + seqlen_q end = start + seqlen_q
seqlen_k = end
input_ids.extend(seq[start:end]) input_ids.extend(seq[start:end])
positions.extend(range(start, end)) positions.extend(range(start, end))
cu_seqlens_q.append(cu_seqlens_q[-1] + seqlen_q) cu_seqlens_q.append(cu_seqlens_q[-1] + seqlen_q)