[V1][PP] Support pp with ray backend in V1 (#1800)

### What this PR does / why we need it?
Support pipeline parallel with ray backend in V1Engine.

Fixes #1751

### Does this PR introduce _any_ user-facing change?
Users could specify ray as distributed backend when inferencing with pp

### How was this patch tested?
CI passed with new added test.


- vLLM version: v0.9.2
- vLLM main:
32142b3c62

---------

Signed-off-by: MengqingCao <cmq0113@163.com>

requirements-dev.txt

@@ -6,7 +6,6 @@ pytest >= 6.0
 pytest-asyncio
 pytest-mock
 lm-eval
-ray
 types-jsonschema
 xgrammar
 zmq
@@ -14,3 +13,5 @@ types-psutil
 pytest-cov
 regex
 sentence_transformers
+ray>=2.47.1
+protobuf==4.25.6

tests/e2e/multicard/test_pipeline_parallel.py

@@ -24,6 +24,7 @@ MODELS = [
 
 TENSOR_PARALLELS = [2]
 PIPELINE_PARALLELS = [2]
+DIST_EXECUTOR_BACKEND = ["mp", "ray"]
 
 prompts = [
     "Hello, my name is",
@@ -34,10 +35,13 @@ prompts = [
 @pytest.mark.parametrize("model", MODELS)
 @pytest.mark.parametrize("tp_size", TENSOR_PARALLELS)
 @pytest.mark.parametrize("pp_size", PIPELINE_PARALLELS)
-def test_models(model: str, tp_size: int, pp_size: int) -> None:
+@pytest.mark.parametrize("distributed_executor_backend", DIST_EXECUTOR_BACKEND)
+def test_models(model: str, tp_size: int, pp_size: int,
+                distributed_executor_backend: str) -> None:
     with VllmRunner(model,
                     tensor_parallel_size=tp_size,
                     pipeline_parallel_size=pp_size,
+                    distributed_executor_backend=distributed_executor_backend,
                     enforce_eager=True,
                     gpu_memory_utilization=0.7) as vllm_model:
         vllm_model.generate_greedy(prompts, 64)

tests/ut/attention/test_attention_v1.py

@@ -400,19 +400,13 @@ class TestAscendAttentionBackendImpl(TestBase):
         layer = self.layer_no_quant
         mock_vanilla_prefill.return_value = MagicMock()
 
-        def mock_tensor(data, device=None, **kwargs):
+        output = self.impl_192.forward(layer,
-            if device == "npu":
+                                       query,
-                return metadata.attn_mask
+                                       key,
-            return torch.tensor(data, **kwargs)
+                                       value,
-
+                                       kv_cache,
-        with patch("torch.tensor", side_effect=mock_tensor):
+                                       metadata,
-            output = self.impl_192.forward(layer,
+                                       trace_flag=False)
-                                           query,
-                                           key,
-                                           value,
-                                           kv_cache,
-                                           metadata,
-                                           trace_flag=False)
 
         mock_vanilla_prefill.assert_called_once()
         assert output.shape == (10, 8 * 192)

vllm_ascend/attention/attention_v1.py

@@ -396,8 +396,10 @@ class AscendAttentionBackendImpl(AttentionImpl):
                 if self.head_size == 192:
                     cu_seqlen_q = [0] + attn_metadata.query_lens.tolist()
                     cu_seqlen_k = [0] + attn_metadata.seq_lens.tolist()
-                    cu_seqlen_q = torch.tensor(cu_seqlen_q, device="npu")
+                    cu_seqlen_q = torch.tensor(cu_seqlen_q,
-                    cu_seqlen_k = torch.tensor(cu_seqlen_k, device="npu")
+                                               device=query.device)
+                    cu_seqlen_k = torch.tensor(cu_seqlen_k,
+                                               device=query.device)
                     cu_seqlen_q = torch.cumsum(cu_seqlen_q, dim=0)
                     cu_seqlen_k = torch.cumsum(cu_seqlen_k, dim=0)
                     max_seqlen_q = torch.max(attn_metadata.query_lens)

vllm_ascend/worker/model_runner_v1.py

@@ -233,7 +233,7 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             self.spec_attn_mask = torch.triu(torch.ones(2048,
                                                         2048,
                                                         dtype=torch.bool),
-                                             diagonal=1).to("npu")
+                                             diagonal=1).to(self.device)
             if get_pp_group().is_last_rank:
                 if self.speculative_config.method == "ngram":
                     self.drafter = NgramProposer(self.vllm_config)
@@ -1120,6 +1120,19 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             input_ids = self.input_ids[:padded_batch_size]
             positions = self.positions[:padded_batch_size]
 
+        if get_pp_group().is_first_rank:
+            intermediate_tensors = None
+        else:
+            assert intermediate_tensors is not None
+            assert self.intermediate_tensors is not None
+            for k, v in intermediate_tensors.items():
+                self.intermediate_tensors[k][:num_input_tokens].copy_(
+                    v[:num_input_tokens], non_blocking=True)
+            intermediate_tensors = IntermediateTensors({
+                k: v[:num_input_tokens]
+                for k, v in self.intermediate_tensors.items()
+            })
+
         # Run forward pass
         with set_forward_context(attn_metadata,
                                  self.vllm_config,