From ae8ac815acc96100faa58c8c183dfc1781e9bcfa Mon Sep 17 00:00:00 2001
From: eugr <eugr@spark.home.eugr.net>
Date: Sat, 9 May 2026 13:45:15 -0700
Subject: [PATCH] Adjusted Qwen3.5-397B recipe to fix OOM issue and lower
 memory requirements

---
 recipes/qwen3.5-397b-int4-autoround.yaml | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/recipes/qwen3.5-397b-int4-autoround.yaml b/recipes/qwen3.5-397b-int4-autoround.yaml
index c688bd1..246e909 100644
--- a/recipes/qwen3.5-397b-int4-autoround.yaml
+++ b/recipes/qwen3.5-397b-int4-autoround.yaml
@@ -18,9 +18,7 @@ container: vllm-node-tf5
 build_args:
   - --tf5
 
-# Mod required to fix ROPE syntax error
 mods:
-  # - mods/fix-qwen3.5-autoround
   - mods/fix-qwen3.5-chat-template
   #- mods/gpu-mem-util-gb
   # - mods/drop-caches
@@ -36,7 +34,6 @@ defaults:
 
 # Environment variables
 env: 
-  #  PYTORCH_CUDA_ALLOC_CONF: "expandable_segments:True" 
   VLLM_MARLIN_USE_ATOMIC_ADD: 1
 
 # The vLLM serve command template