update fa full_cuda_graph support

vllm/config.py

@@ -4135,7 +4135,7 @@ class CompilationConfig:
     are always used, it can set this to False. Otherwise, it should
     set this to True, and the compiler will copy the input to an
     internally managed buffer. Default is False."""
-    full_cuda_graph: bool = False
+    full_cuda_graph: bool = True
     """whether to use a full cuda graph for the entire forward pass rather than
     splitting certain operations such as attention into subgraphs. Thus this
     flag cannot be used together with splitting_ops. This may provide

vllm/v1/attention/backends/flash_attn.py

@@ -151,7 +151,7 @@ def _get_sliding_window_configs(
 
 class FlashAttentionMetadataBuilder(
         AttentionMetadataBuilder[FlashAttentionMetadata]):
-    full_cudagraph_supported: ClassVar[bool] = get_flash_attn_version() == 3
+    full_cudagraph_supported: ClassVar[bool] = get_flash_attn_version() == 3 or current_platform.is_rocm() 
 
     def __init__(self, kv_cache_spec: AttentionSpec, vllm_config: VllmConfig,
                  device: torch.device):
@@ -172,7 +172,8 @@ class FlashAttentionMetadataBuilder(
         self.max_num_splits = 0  # No upper bound on the number of splits.
         self.aot_schedule = (get_flash_attn_version() == 3)
         self.use_full_cuda_graph = self.compilation_config.full_cuda_graph
-        if not current_platform.is_rocm() and self.use_full_cuda_graph:
+        if self.use_full_cuda_graph:
+            if not current_platform.is_rocm():
                 if not self.aot_schedule:
                     raise ValueError(
                         "AoT scheduling is required for full cuda graph.")
@@ -325,7 +326,7 @@ class FlashAttentionMetadataBuilder(
             scheduler_metadata = self.scheduler_metadata[:n]
 
         max_num_splits = 0
-        if (not current_platform.is_rocm() and self.use_full_cuda_graph
+        if (self.use_full_cuda_graph
                 and num_actual_tokens <= self.max_cudagraph_size):
             # NOTE(woosuk): Setting num_splits > 1 may increase the memory
             # usage, because the intermediate buffers of size [num_splits,

vllm/v1/worker/gpu_model_runner.py

@@ -2548,7 +2548,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                 self.device,
             )
 
-            if (not current_platform.is_rocm() and self.full_cuda_graph
+            if (self.full_cuda_graph
                     and not attn_metadata_builder_i.full_cudagraph_supported):
                 raise ValueError(
                     f"Full CUDAGraph not supported for "