[ROCm] To reduce the compiling time when using torch compile. (#10559)

python/sglang/srt/model_executor/cuda_graph_runner.py

@@ -53,7 +53,9 @@ from sglang.srt.two_batch_overlap import TboCudaGraphRunnerPlugin
 from sglang.srt.utils import (
     empty_context,
     get_available_gpu_memory,
+    get_bool_env_var,
     get_device_memory_capacity,
+    is_hip,
     log_info_on_rank0,
     require_attn_tp_gather,
     require_gathered_buffer,
@@ -61,6 +63,8 @@ from sglang.srt.utils import (
     require_mlp_tp_gather,
 )
 
+_is_hip = is_hip()
+
 logger = logging.getLogger(__name__)
 
 if TYPE_CHECKING:
@@ -137,7 +141,7 @@ def patch_model(
                 mode=os.environ.get(
                     "SGLANG_TORCH_COMPILE_MODE", "max-autotune-no-cudagraphs"
                 ),
-                dynamic=False,
+                dynamic=_is_hip and get_bool_env_var("SGLANG_TORCH_DYNAMIC_SHAPE"),
             )
         else:
             yield model.forward