allow for cuda graphs again

2026-05-16 19:23:41 +00:00
parent d15c43294b
commit da31ce7e1a
1 changed files with 3 additions and 2 deletions
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -15,9 +15,10 @@ services:
      - --trust-remote-code
      - --enable-expert-parallel
      - --tensor-parallel-size=8
-      - --enforce-eager
-      #- --compilation-config
+      #- --enforce-eager
+      - --compilation-config
      #- '{"cudagraph_mode": "FULL_DECODE_ONLY", "custom_ops": ["all"], "cudagraph_capture_sizes": [1, 2, 4, 8], "max_cudagraph_capture_size": 8}'
+      - '{"cudagraph_mode":"FULL_AND_PIECEWISE", "custom_ops":["all"]}'
      - --tokenizer-mode=deepseek_v4
      - --host=0.0.0.0
      - --port=8000