sirius/gcloud-lab
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

dial in vllm settings

Browse source
This commit is contained in:
sirius0xdev 2026-04-19 23:41:15 +00:00
parent 28e701e27f
commit ee3cf7c7ee
1 changed files with 4 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
apps/base/customer1/openclaw/vllm-gemma.yaml
View file

@ -36,16 +36,16 @@ spec:
- --host=0.0.0.0 - --host=0.0.0.0
- --port=8000 - --port=8000
- --tensor-parallel-size=1 - --tensor-parallel-size=1
- --gpu-memory-utilization=0.78 # ← Critical: lower than default 0.90 - --gpu-memory-utilization=0.98 # ← Critical: lower than default 0.90
- --max-model-len=16384 # Start conservative (you can raise to 32768 later) - --max-model-len=24576 # Start conservative (you can raise to 32768 later)
- --max-num-batched-tokens=8192 - --max-num-batched-tokens=8192
- --max-num-seqs=4 # Lower for single-user / low-concurrency - --max-num-seqs=16 # Lower for single-user / low-concurrency
- --enforce-eager # Disables CUDA graphs → big memory saver at startup - --enforce-eager # Disables CUDA graphs → big memory saver at startup
- --disable-custom-all-reduce - --disable-custom-all-reduce
- --enable-auto-tool-choice
- --trust-remote-code - --trust-remote-code
- --dtype=auto - --dtype=auto
- --tool-call-parser qwen3_xml - --tool-call-parser qwen3_xml
- --reasoning-parser qwen3
- --enable-prefix-caching - --enable-prefix-caching
##- --limit-mm-per-prompt='{"image"=1, "video"=0}' # Optional: limit vision if not heavily using images yet ##- --limit-mm-per-prompt='{"image"=1, "video"=0}' # Optional: limit vision if not heavily using images yet
ports: ports: