[V1] [Hybrid] Enable Full CUDA Graph (decode-only) for Mamba layers

tests/models/language/generation/test_hybrid.py

@@ -387,3 +387,63 @@ def test_distributed_correctness(
         name_0="vllm_tp_1",
         name_1="vllm_tp_2",
     )
+
+
+@pytest.mark.parametrize("model", ["Zyphra/Zamba2-1.2B-instruct"])
+@pytest.mark.parametrize("max_tokens", [64])
+@pytest.mark.parametrize("num_logprobs", [5])
+def test_full_cuda_graph(
+    hf_runner,
+    vllm_runner,
+    example_prompts,
+    monkeypatch,
+    model: str,
+    max_tokens: int,
+    num_logprobs: int,
+) -> None:
+
+    try:
+        model_info = HF_EXAMPLE_MODELS.find_hf_info(model)
+        model_info.check_available_online(on_fail="skip")
+        model_info.check_transformers_version(on_fail="skip")
+    except ValueError:
+        pass
+
+    with hf_runner(model) as hf_model:
+        if model not in HF_UNSUPPORTED_MODELS:
+            hf_outputs = hf_model.generate_greedy_logprobs_limit(
+                example_prompts, max_tokens, num_logprobs)
+        else:
+            hf_outputs = None
+
+    with vllm_runner(model, max_num_seqs=MAX_NUM_SEQS) as vllm_model:
+        vllm_v0_outputs = vllm_model.generate_greedy_logprobs(
+            example_prompts, max_tokens, num_logprobs)
+
+    with monkeypatch.context() as m:
+        m.setenv("VLLM_USE_V1", "1")
+        if model in HYBRID_MODELS:
+            # required due to reorder_batch behaviour
+            m.setenv("VLLM_ATTENTION_BACKEND", "FLASHINFER")
+        with vllm_runner(model,
+                         max_num_seqs=MAX_NUM_SEQS,
+                         compilation_config={'full_cuda_graph': True},
+                         enable_prefix_caching=False) as vllm_model:
+            vllm_v1_outputs = vllm_model.generate_greedy_logprobs(
+                example_prompts, max_tokens, num_logprobs)
+
+    if hf_outputs is not None:
+        check_logprobs_close(
+            outputs_0_lst=hf_outputs,
+            outputs_1_lst=vllm_v0_outputs,
+            name_0="hf",
+            name_1="vllm-v0",
+        )
+
+    ref_outputs = hf_outputs if hf_outputs is not None else vllm_v0_outputs
+    check_logprobs_close(
+        outputs_0_lst=ref_outputs,
+        outputs_1_lst=vllm_v1_outputs,
+        name_0="hf" if hf_outputs is not None else "vllm-v0",
+        name_1="vllm-v1",
+    )

vllm/v1/attention/backends/mamba_attn.py

@@ -8,7 +8,8 @@
 
 from vllm.attention.backends.abstract import AttentionBackend
 from vllm.config import VllmConfig
-from vllm.v1.attention.backends.utils import (AttentionMetadataBuilder,
+from vllm.v1.attention.backends.utils import (AttentionCGSupport,
+                                              AttentionMetadataBuilder,
                                               CommonAttentionMetadata,
                                               split_decodes_and_prefills)
 from vllm.v1.kv_cache_interface import AttentionSpec, MambaSpec
@@ -82,6 +83,8 @@ class Mamba2AttentionMetadata:
 
 class Mamba2AttentionMetadataBuilder(
         AttentionMetadataBuilder[Mamba2AttentionMetadata]):
+    attn_cudagraph_support: ClassVar[AttentionCGSupport] = \
+        AttentionCGSupport.PURE_DECODE_ONLY
 
     reorder_batch_threshold: ClassVar[int] = 1
 
@@ -160,3 +163,23 @@ def build(self,
             state_indices_tensor=state_indices_tensor,
         )
         return attn_metadata
+
+    def build_for_cudagraph_capture(
+            self, common_attn_metadata: CommonAttentionMetadata):
+        """
+        This method builds the metadata for full cudagraph capture.
+        Currently, only decode is supported for full cudagraphs with Mamba.
+        """
+        m = common_attn_metadata
+
+        assert m.num_reqs == m.num_actual_tokens, \
+            "Mamba only supports decode-only full CUDAGraph capture. " \
+            "Make sure all cudagraph capture sizes <= max_num_seq."
+
+        m.max_query_len = 1  # decode-only
+
+        return self.build(0, m)
+
+    def can_run_in_cudagraph(
+            self, common_attn_metadata: CommonAttentionMetadata) -> bool:
+        return common_attn_metadata.max_query_len == 1