xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 08:04:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] Fix seq_lens for graph capture (#23175)

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon 2025-08-19 03:58:16 -07:00 committed by GitHub
parent 2c3f557f08
commit 40f26734b9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/v1/worker/gpu_model_runner.py
View File

@ -2317,15 +2317,13 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
# If force_attention is True, we always capture attention. Otherwise,
# it only happens for cudagraph_runtime_mode=FULL.
if force_attention or cudagraph_runtime_mode == \
CUDAGraphMode.FULL:
if force_attention or cudagraph_runtime_mode == CUDAGraphMode.FULL:
attn_metadata = {}
# Make sure max_model_len is used at the graph capture time.
self.seq_lens_np[:num_reqs] = self.max_model_len
self.seq_lens_np[num_reqs:] = 0
self.seq_lens[:num_reqs].copy_(self.seq_lens_cpu[:num_reqs],
non_blocking=True)
self.seq_lens.copy_(self.seq_lens_cpu, non_blocking=True)
for kv_cache_group_id, kv_cache_group_spec in enumerate(
self.kv_cache_config.kv_cache_groups):