xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-24 00:35:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix Whisper crash caused by invalid`` max_num_batched_tokens`` config (#17853)

Browse Source 
Signed-off-by: inkcherry <mingzhi.liu@intel.com>
This commit is contained in:
inkcherry 2025-05-09 17:16:26 +08:00 committed by GitHub
parent 6e4a93e3f7
commit 5b2dcbf0b8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 14 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
vllm/config.py
View File

@ -2050,6 +2050,13 @@ class SchedulerConfig:
_MULTIMODAL_MODEL_MAX_NUM_BATCHED_TOKENS,
)
# When using default settings,
# Ensure max_num_batched_tokens does not exceed model limit.
# Some models (e.g., Whisper) have embeddings tied to max length.
self.max_num_batched_tokens = min(
self.max_num_seqs * self.max_model_len,
self.max_num_batched_tokens)
self.max_num_encoder_input_tokens = self.max_num_batched_tokens
self.encoder_cache_size = self.max_num_batched_tokens
@ -2090,6 +2097,13 @@ class SchedulerConfig:
"be greater than or equal to max_num_seqs "
f"({self.max_num_seqs}).")
if self.max_num_batched_tokens > self.max_num_seqs * self.max_model_len:
logger.warning(
"max_num_batched_tokens (%d) exceeds max_num_seqs"
"* max_model_len (%d). This may lead to unexpected behavior.",
self.max_num_batched_tokens,
self.max_num_seqs * self.max_model_len)
if self.num_lookahead_slots < 0:
raise ValueError(
"num_lookahead_slots "