software-engineering/spark-vllm-docker
2
0
Fork 0
Code Issues Actions Packages Projects Releases Wiki Activity

switching gpt-oss-120b to solo only for now

Browse Source
This commit is contained in:
Eugene Rakhmatulin
2026-04-01 10:27:50 -07:00
parent 27eb35f08d
commit 12caec228e
1 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
recipes/openai-gpt-oss-120b.yaml
View File

@@ -11,6 +11,9 @@ model: openai/gpt-oss-120b
# Container image to use # Container image to use
container: vllm-node-mxfp4 container: vllm-node-mxfp4
# Only solo now
solo_only: true
# Build arguments for build-and-copy.sh # Build arguments for build-and-copy.sh
build_args: build_args:
- --exp-mxfp4 - --exp-mxfp4
@@ -22,7 +25,7 @@ mods: []
defaults: defaults:
port: 8000 port: 8000
host: 0.0.0.0 host: 0.0.0.0
tensor_parallel: 2 tensor_parallel: 1
gpu_memory_utilization: 0.70 gpu_memory_utilization: 0.70
max_num_batched_tokens: 8192 max_num_batched_tokens: 8192
@@ -37,8 +40,6 @@ command: |
--tool-call-parser openai \ --tool-call-parser openai \
--reasoning-parser openai_gptoss \ --reasoning-parser openai_gptoss \
--enable-auto-tool-choice \ --enable-auto-tool-choice \
--tensor-parallel-size {tensor_parallel} \
--distributed-executor-backend ray \
--gpu-memory-utilization {gpu_memory_utilization} \ --gpu-memory-utilization {gpu_memory_utilization} \
--enable-prefix-caching \ --enable-prefix-caching \
--load-format fastsafetensors \ --load-format fastsafetensors \