Main2main upgrade vllm commit to 03 19 17:00 (#7478)

### What this PR does / why we need it? Upgrade vllm commit to 2026.03.19. 1.Fix socket removed from StatelessProcessGroup. Upstream vLLM PR [#36330](https://github.com/vllm-project/vllm/pull/36330) ("elastic_ep: Fix stateless group port races") refactored StatelessProcessGroup and removed the socket: socket.socket | None field. The socket ownership was moved to a new create_tcp_store() helper instead of being stored as a field on the dataclass. 2.fix `virtual_engine` parameter removed from `set_forward_context(). Upstream [V0 Deprecation] Deprecate virtual engine [#37195](https://github.com/vllm-project/vllm/pull/37195) ### Does this PR introduce _any_ user-facing change? NA ### How was this patch tested? NA - vLLM version: v0.17.0 - vLLM main: 8b6325758c --------- Signed-off-by: leo-pony <nengjunma@outlook.com>
2026-03-23 16:25:57 +08:00
parent caa71e50ca
commit 8e2c59e1ee
12 changed files with 28 additions and 20 deletions
--- a/.github/workflows/bot_pr_create.yaml
+++ b/.github/workflows/bot_pr_create.yaml
@@ -37,7 +37,7 @@ jobs:
    steps:
      - name: Get vLLM version
        run: |
-          VLLM_COMMIT=8b6325758cce5f9c36d38f2462edbd368b97a07c
+          VLLM_COMMIT=6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209
          echo "VLLM_COMMIT=https://github.com/vllm-project/vllm/commit/$VLLM_COMMIT" >> "$GITHUB_ENV"

      - name: Checkout repository
--- a/.github/workflows/dockerfiles/Dockerfile.lint
+++ b/.github/workflows/dockerfiles/Dockerfile.lint
@@ -27,7 +27,7 @@ RUN apt-get update -y && \

 ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
 # For lint purpose, actually we need make a main2main matching.
-ARG VLLM_COMMIT=8b6325758cce5f9c36d38f2462edbd368b97a07c
+ARG VLLM_COMMIT=6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209
 RUN git clone $VLLM_REPO /vllm-workspace/vllm && \
    cd /vllm-workspace/vllm && \
    git checkout $VLLM_COMMIT
--- a/.github/workflows/pr_test_full.yaml
+++ b/.github/workflows/pr_test_full.yaml
@@ -75,7 +75,7 @@ jobs:
    name: e2e-full
    strategy:
      matrix:
-        vllm_version: [8b6325758cce5f9c36d38f2462edbd368b97a07c, v0.18.0]
+        vllm_version: [6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209, v0.18.0]
    needs: [changes]
    if: ${{ needs.changes.outputs.e2e_tracker == 'true' || needs.changes.outputs.e2e_tracker == true }}
    uses: ./.github/workflows/_e2e_test.yaml
--- a/.github/workflows/pr_test_light.yaml
+++ b/.github/workflows/pr_test_light.yaml
@@ -41,7 +41,7 @@ jobs:
  lint:
    uses: ./.github/workflows/_pre_commit.yml
    with:
-      vllm: 8b6325758cce5f9c36d38f2462edbd368b97a07c
+      vllm: 6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209
  changes:
    runs-on: linux-aarch64-a2b3-0
    outputs:
@@ -90,7 +90,7 @@ jobs:
    if: ${{ needs.lint.result == 'success' && (needs.changes.outputs.e2e_tracker == 'true' || needs.changes.outputs.ut_tracker == 'true') }}
    strategy:
      matrix:
-        vllm_version: [8b6325758cce5f9c36d38f2462edbd368b97a07c, v0.18.0]
+        vllm_version: [6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209, v0.18.0]
    uses: ./.github/workflows/_unit_test.yaml
    with:
      vllm: ${{ matrix.vllm_version }}
@@ -102,7 +102,7 @@ jobs:
    name: e2e-light
    strategy:
      matrix:
-        vllm_version: [8b6325758cce5f9c36d38f2462edbd368b97a07c, v0.18.0]
+        vllm_version: [6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209, v0.18.0]
    # Note (yikun): If CI resource are limited we can split job into two chain jobs
    needs: [lint, changes]
    # only trigger e2e test after lint passed and the change is e2e related with pull request.
--- a/.github/workflows/schedule_codecov_refresh.yaml
+++ b/.github/workflows/schedule_codecov_refresh.yaml
@@ -33,7 +33,7 @@ jobs:
    name: refresh codecov
    strategy:
      matrix:
-        vllm_version: [8b6325758cce5f9c36d38f2462edbd368b97a07c]
+        vllm_version: [6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209]
    uses: ./.github/workflows/_unit_test.yaml
    with:
      vllm: ${{ matrix.vllm_version }}
--- a/docs/source/community/versioning_policy.md
+++ b/docs/source/community/versioning_policy.md
@@ -59,7 +59,7 @@ For main branch of vLLM Ascend, we usually make it compatible with the latest vL

 | vLLM Ascend | vLLM         | Python           | Stable CANN | PyTorch/torch_npu  |
 |-------------|--------------|------------------|-------------|--------------------|
-|     main    | 8b6325758cce5f9c36d38f2462edbd368b97a07c, v0.18.0 tag | >= 3.10, < 3.12   | 8.5.0 | 2.9.0 / 2.9.0 |
+|     main    | 6a9cceb219fcbd6b1eb540ddfdc77ec160f0e209, v0.18.0 tag | >= 3.10, < 3.12   | 8.5.0 | 2.9.0 / 2.9.0 |

 ## Release cadence

--- a/tests/ut/distributed/device_communicators/test_pyhccl.py
+++ b/tests/ut/distributed/device_communicators/test_pyhccl.py
@@ -6,6 +6,7 @@ from vllm.distributed.utils import StatelessProcessGroup
 from tests.ut.base import TestBase
 from vllm_ascend.distributed.device_communicators.pyhccl import \
    PyHcclCommunicator
+from vllm_ascend.utils import ACL_FORMAT_FRACTAL_NZ, vllm_version_is


 class MockHcclLib:
@@ -45,10 +46,15 @@ class TestPyHcclCommunicator(TestBase):
    @patch("vllm_ascend.utils.current_stream",
           return_value=MagicMock(npu_stream=5678))
    def test_stateless_group(self, *_):
-        group = StatelessProcessGroup(rank=3,
-                                      world_size=4,
-                                      store=None,
-                                      socket=None)
+        if vllm_version_is("0.18.0"):
+            group = StatelessProcessGroup(rank=3,
+                                        world_size=4,
+                                        store=None,
+                                        socket=None)
+        else:
+            group = StatelessProcessGroup(rank=3,
+                                        world_size=4,
+                                        store=None)

        comm = PyHcclCommunicator(group=group, device=3)

--- a/vllm_ascend/ascend_forward_context.py
+++ b/vllm_ascend/ascend_forward_context.py
@@ -19,6 +19,7 @@ from vllm_ascend.utils import (
    is_drafter_moe_model,
    is_moe_model,
    speculative_enable_dispatch_gmm_combine_decode,
+    vllm_version_is,
 )


@@ -53,13 +54,14 @@ def set_ascend_forward_context(
    forward_context_kwargs = {
        "attn_metadata": attn_metadata,
        "vllm_config": vllm_config,
-        "virtual_engine": virtual_engine,
        "num_tokens": num_tokens,
        "num_tokens_across_dp": num_tokens_across_dp,
        "cudagraph_runtime_mode": aclgraph_runtime_mode,
        "batch_descriptor": batch_descriptor,
        "skip_compiled": skip_compiled,
    }
+    if vllm_version_is("0.18.0"):
+        forward_context_kwargs["virtual_engine"] = virtual_engine

    with set_forward_context(**forward_context_kwargs):
        forward_context = get_forward_context()
--- a/vllm_ascend/ops/mla.py
+++ b/vllm_ascend/ops/mla.py
@@ -33,7 +33,7 @@ from vllm.v1.attention.backend import AttentionMetadata  # type: ignore

 from vllm_ascend.ascend_config import get_ascend_config
 from vllm_ascend.ascend_forward_context import _EXTRA_CTX
-from vllm_ascend.utils import is_vl_model, parse_layer_idx
+from vllm_ascend.utils import is_vl_model, parse_layer_idx, vllm_version_is


 class IndexerWrapper(nn.Module):
@@ -183,7 +183,7 @@ def mla_forward(
        attn_metadata = forward_context.attn_metadata[self.mla_attn.layer_name]
    else:
        attn_metadata = forward_context.attn_metadata
-    kv_cache = self.mla_attn.kv_cache[forward_context.virtual_engine]
+    kv_cache = self.mla_attn.kv_cache[forward_context.virtual_engine if vllm_version_is("0.18.0") else 0]
    self.mla_attn.impl.forward(
        self.mla_attn.layer_name, hidden_states, kv_cache, attn_metadata, need_gather_q_kv, output
    )
--- a/vllm_ascend/patch/worker/patch_qwen3_5.py
+++ b/vllm_ascend/patch/worker/patch_qwen3_5.py
@@ -29,7 +29,7 @@ from vllm.v1.attention.backends.utils import PAD_SLOT_ID
 from vllm_ascend.attention.utils import maybe_save_kv_layer_to_connector
 from vllm_ascend.ops.triton.fla.sigmoid_gating import fused_sigmoid_gating_delta_rule_update
 from vllm_ascend.ops.triton.fused_gdn_gating import fused_gdn_gating_patch
-from vllm_ascend.utils import enable_sp
+from vllm_ascend.utils import enable_sp, vllm_version_is


 class AscendQwen3_5GatedDeltaNet(Qwen3_5GatedDeltaNet):
@@ -66,7 +66,7 @@ class AscendQwen3_5GatedDeltaNet(Qwen3_5GatedDeltaNet):
        non_spec_token_indx = attn_metadata.non_spec_token_indx
        spec_state_indices_tensor = attn_metadata.spec_state_indices_tensor  # noqa: E501
        non_spec_state_indices_tensor = attn_metadata.non_spec_state_indices_tensor  # noqa: E501
-        self_kv_cache = self.kv_cache[forward_context.virtual_engine]
+        self_kv_cache = self.kv_cache[forward_context.virtual_engine if vllm_version_is("0.18.0") else 0]
        conv_state = self_kv_cache[0].transpose(-1, -2)
        ssm_state = self_kv_cache[1]
        num_actual_tokens = attn_metadata.num_actual_tokens
--- a/vllm_ascend/patch/worker/patch_qwen3_next.py
+++ b/vllm_ascend/patch/worker/patch_qwen3_next.py
@@ -32,7 +32,7 @@ from vllm.v1.attention.backends.utils import PAD_SLOT_ID
 from vllm_ascend.attention.utils import maybe_save_kv_layer_to_connector
 from vllm_ascend.ops.triton.fla.fused_qkvzba_split_reshape import fused_qkvzba_split_reshape_cat
 from vllm_ascend.ops.triton.fused_gdn_gating import fused_gdn_gating_patch
-from vllm_ascend.utils import enable_sp
+from vllm_ascend.utils import enable_sp, vllm_version_is


 class AscendQwen3Next_GatedDeltaNet(Qwen3NextGatedDeltaNet):
@@ -124,7 +124,7 @@ class AscendQwen3Next_GatedDeltaNet(Qwen3NextGatedDeltaNet):
        non_spec_token_indx = attn_metadata.non_spec_token_indx
        spec_state_indices_tensor = attn_metadata.spec_state_indices_tensor  # noqa: E501
        non_spec_state_indices_tensor = attn_metadata.non_spec_state_indices_tensor  # noqa: E501
-        self_kv_cache = self.kv_cache[forward_context.virtual_engine]
+        self_kv_cache = self.kv_cache[forward_context.virtual_engine if vllm_version_is("0.18.0") else 0]
        conv_state = self_kv_cache[0].transpose(-1, -2)
        ssm_state = self_kv_cache[1]
        num_actual_tokens = attn_metadata.num_actual_tokens
--- a/vllm_ascend/platform.py
+++ b/vllm_ascend/platform.py
@@ -551,7 +551,7 @@ class NPUPlatform(Platform):
        attn_metadata: dict[str, Any],
        vllm_config: VllmConfig,
        dp_metadata,
-        virtual_engine: int = 0,
+        virtual_engine: int = 0,  # ToDo:: Remove me when upgrade to vllm 0.19.0 from 0.18.0
        num_tokens: int = 0,
        num_tokens_across_dp: torch.Tensor | None = None,
        cudagraph_runtime_mode=None,