fix(docker): clarify GPU smoke completion and exit status

Author: KraHsu

docker/README.md

@@ -75,7 +75,9 @@ docker compose run --rm -T gpu-check
 
 This checks BF16, compiled forward/backward, DeepSpeed FusedAdam and multi-GPU
 NCCL all-reduce with networking disabled. It requires the cache/output mounts,
-but no weights or dataset. Follow it with real inference or training.
+but no weights or dataset. Success requires the entire command to exit with
+code 0: worker computation messages alone do not rule out a shutdown failure.
+Follow it with real inference or training.
 
 ## Serve
 

docker/gpu_smoke.py

@@ -64,7 +64,7 @@ def operation(value):
         finally:
             dist.destroy_process_group()
     torch.cuda.synchronize()
-    print("GPU smoke passed.", flush=True)
+    print(f"Rank {rank}: GPU computation checks completed.", flush=True)
 
 
 if __name__ == "__main__":