owen/worker-vllm
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #265 from runpod-workers/fix/zero-max-model-num_batches
Release / release (push) Waiting to run

Browse Source 
fix: check for zero param and set to None
This commit is contained in:
chrisvela
2026-02-13 15:26:06 -06:00
committed by GitHub
parent b749aa5718 e705c9494b
commit 1606cff557
1 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/engine_args.py
+7
View File
@@ -288,6 +288,13 @@ def get_engine_args():
# Set max_num_batched_tokens to max_model_len for unlimited batching. # Set max_num_batched_tokens to max_model_len for unlimited batching.
# vLLM defaults max_num_batched_tokens to 2048 when None, which is too low. # vLLM defaults max_num_batched_tokens to 2048 when None, which is too low.
if args.get("max_model_len") == 0:
args["max_model_len"] = None
if args.get("max_num_batched_tokens") == 0:
args["max_num_batched_tokens"] = None
if args.get("max_num_batched_tokens") is None: if args.get("max_num_batched_tokens") is None:
max_model_len = args.get("max_model_len") max_model_len = args.get("max_model_len")
if max_model_len is None: if max_model_len is None: