x

Dockerfile

@@ -4,6 +4,9 @@ FROM docker.1ms.run/kyuz0/vllm-therock-gfx1201:latest
 ENV USE_DEFAULT_MODEL=true
 ENV LOCAL_MODEL_DIR=/opt/model
 
+# Ensure latest Transformers model definitions are available
+RUN /opt/venv/bin/pip install --no-cache-dir --upgrade "git+https://github.com/huggingface/transformers.git"
+
 # Create necessary directories
 RUN mkdir -p /opt/script /opt/model /config
 

config.json

@@ -99,6 +99,18 @@
       "served_model_name": "GLM-4.7-Flash-AWQ",
       "hf_model_id": "THUDM/GLM-4.7-Flash-AWQ"
     },
+    "Qwen3.5-27B-FP8": {
+      "ctx": "32768",
+      "trust_remote": true,
+      "valid_tp": [1, 2],
+      "max_num_seqs": "64",
+      "max_tokens": "32768",
+      "gpu_util": "0.98",
+      "tool_call_parser": "qwen3_xml",
+      "enable_auto_tool_choice": true,
+      "served_model_name": "Qwen3.5-27B-FP8",
+      "hf_model_id": "RedHatAI/Qwen3.5-27B-FP8-dynamic"
+    },
     "Qwen3.5-35B-A3B-GPTQ-Int4": {
       "ctx": "32768",
       "trust_remote": true,