fix: revert to eager mode — CUDA graphs OOM with 175GB model

CUDA graph capture needs extra memory on top of the model weights.
With 175GB model on 178GB GPUs, there's no room.

Going back to --enforce-eager with 10-min RPC timeout. The first
inference request will be slow (2-3 min JIT compilation) but won't
crash. Subsequent requests are fast.

CUDA graph mode requires either more GPU memory or a smaller model.

docker-compose.yml

@@ -15,8 +15,7 @@ services:
       - --trust-remote-code
       - --enable-expert-parallel
       - --tensor-parallel-size=8
-      - --compilation-config
-      - '{"cudagraph_mode":"FULL_AND_PIECEWISE", "custom_ops":["all"]}'
+      - --enforce-eager
       - --tokenizer-mode=deepseek_v4
       - --host=0.0.0.0
       - --port=8000