x

Dockerfile

@@ -7,6 +7,9 @@ ENV LOCAL_MODEL_DIR=/opt/model
 # Create necessary directories
 RUN mkdir -p /opt/script /opt/model /config
 
+# Update transformers to support qwen3_5_moe architecture
+RUN pip install --upgrade transformers
+
 # Copy scripts to /opt/script
 COPY scripts/start_vllm.py /opt/script/start-vllm
 COPY benchmarks/run_vllm_bench.py /opt/script/run_vllm_bench.py

config.json

@@ -1,5 +1,5 @@
 {
-  "default_model": "Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit",
+  "default_model": "Qwen3.5-35B-A3B-GPTQ-Int4",
   "models": {
     "gpt-oss-20b": {
       "ctx": "32768",