diff --git a/Dockerfile b/Dockerfile
index fa70850..ac76c9f 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,5 +1,4 @@
-# FROM docker.1ms.run/vllm/vllm-openai-rocm:latest
-FROM vllm/vllm-openai-rocm:nightly
+FROM vllm/vllm-openai-rocm:latest
 
 WORKDIR /workspace
 
diff --git a/README.md b/README.md
index 2dfc4c7..23ef15c 100644
--- a/README.md
+++ b/README.md
@@ -124,7 +124,7 @@ curl -X POST "http://localhost:8001/v1/chat/completions" \
 
 ## 常见故障排查
 
-- 报错 `Model architectures ['Qwen3_5MoeForConditionalGeneration'] are not supported for now` 时，在对应 profile 增加 `model_impl=transformers`，并使用当前 Dockerfile 重建镜像后再启动。
+- 报错 `Model architectures ['Qwen3_5MoeForConditionalGeneration'] are not supported for now` 或 `The Transformers implementation ... is not compatible with vLLM` 时，说明当前 vLLM 栈与该模型架构不兼容，需切换到兼容模型或改用其他推理后端。
 - 报错 `model config (gptq) does not match quantization argument (gptq_marlin)` 时，将该模型配置改为 `dtype=float16` 且 `quantization=gptq`。
 - 报错 `RPC call to sample_tokens timed out` 或出现 `GPU core dump` 时，先下调模型配置为更稳参数：`ctx=32768`、`max_num_seqs=4`、`max_tokens=2048`、`gpu_util=0.90`，并开启 `enforce_eager=true`。
 - 若模型目录存在但仍加载失败，检查挂载路径是否为 `/opt/model:/opt/model:ro`，并确认容器内可见模型文件。
diff --git a/config.json b/config.json
index 1f828c0..f3c6e72 100644
--- a/config.json
+++ b/config.json
@@ -73,7 +73,6 @@
         "local_path": "Qwen3.5-35B-A3B-GPTQ-Int4",
         "dtype": "float16",
         "quantization": "moe_wna16",
-        "model_impl": "transformers",
         "ctx": "32768",
         "trust_remote": true,
         "valid_tp": [1, 2],