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

Adjusted Qwen3.5-397B recipe to fix OOM issue and lower memory requirements

Browse Source
This commit is contained in:
eugr
2026-05-09 13:45:15 -07:00
parent 83a680c87b
commit ae8ac815ac
1 changed files with 0 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
recipes/qwen3.5-397b-int4-autoround.yaml
View File

@@ -18,9 +18,7 @@ container: vllm-node-tf5
build_args: build_args:
- --tf5 - --tf5
# Mod required to fix ROPE syntax error
mods: mods:
# - mods/fix-qwen3.5-autoround
- mods/fix-qwen3.5-chat-template - mods/fix-qwen3.5-chat-template
#- mods/gpu-mem-util-gb #- mods/gpu-mem-util-gb
# - mods/drop-caches # - mods/drop-caches
@@ -36,7 +34,6 @@ defaults:
# Environment variables # Environment variables
env: env:
# PYTORCH_CUDA_ALLOC_CONF: "expandable_segments:True"
VLLM_MARLIN_USE_ATOMIC_ADD: 1 VLLM_MARLIN_USE_ATOMIC_ADD: 1
# The vLLM serve command template # The vLLM serve command template