fix import vllm_version_is

Signed-off-by: Icey <1790571317@qq.com>

Author: wxsIcey

.github/workflows/format_pr_body.yaml

@@ -36,7 +36,7 @@ jobs:
 
       - name: Get vLLM version
         run: |
-          VLLM_COMMIT=17c540a993af88204ad1b78345c8a865cf58ce44
+          VLLM_COMMIT=9fce7bee745230d61c60ad467966790553b0ba48
           echo "VLLM_COMMIT=https://github.com/vllm-project/vllm/commit/$VLLM_COMMIT" >> $GITHUB_ENV
 
       - name: Checkout repository

.github/workflows/vllm_ascend_test.yaml

@@ -42,7 +42,7 @@ jobs:
   lint:
     uses: ./.github/workflows/pre-commit.yml
     with:
-      vllm: 17c540a993af88204ad1b78345c8a865cf58ce44
+      vllm: 9fce7bee745230d61c60ad467966790553b0ba48
 
   changes:
     runs-on: ubuntu-latest
@@ -83,7 +83,7 @@ jobs:
         VLLM_USE_MODELSCOPE: True
     strategy:
       matrix:
-        vllm_version: [17c540a993af88204ad1b78345c8a865cf58ce44, v0.11.0]
+        vllm_version: [9fce7bee745230d61c60ad467966790553b0ba48, v0.11.0]
     steps:
       - name: Install packages
         run: |
@@ -146,7 +146,7 @@ jobs:
     name: e2e-light
     strategy:
       matrix:
-        vllm_version: [17c540a993af88204ad1b78345c8a865cf58ce44, v0.11.0]
+        vllm_version: [9fce7bee745230d61c60ad467966790553b0ba48, v0.11.0]
     # 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.

.github/workflows/vllm_ascend_test_full.yaml

@@ -69,7 +69,7 @@ jobs:
     name: e2e-full
     strategy:
       matrix:
-        vllm_version: [17c540a993af88204ad1b78345c8a865cf58ce44, v0.11.0]
+        vllm_version: [9fce7bee745230d61c60ad467966790553b0ba48, v0.11.0]
     needs: [changes]
     if: ${{ needs.changes.outputs.e2e_tracker == 'true' }}
     uses: ./.github/workflows/_e2e_test.yaml

vllm_ascend/attention/attention_v1.py

@@ -26,7 +26,7 @@
                                               AttentionLayer, AttentionType)
 from vllm.config import VllmConfig
 from vllm.forward_context import ForwardContext, get_forward_context
-from vllm.utils import cdiv, direct_register_custom_op
+from vllm.utils import cdiv
 from vllm.v1.attention.backends.utils import AttentionCGSupport
 from vllm.v1.core.sched.output import SchedulerOutput
 from vllm.v1.kv_cache_interface import AttentionSpec
@@ -38,10 +38,16 @@
                                                update_graph_params_workspaces)
 from vllm_ascend.ops.attention import vanilla_chunked_prefill
 from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_NZ, aligned_16, is_310p,
-                               nd_to_nz_2d, nd_to_nz_spec, version_check)
+                               nd_to_nz_2d, nd_to_nz_spec, version_check,
+                               vllm_version_is)
 
 from ..utils import weak_ref_tensors
 
+if vllm_version_is("0.11.0"):
+    from vllm.utils import direct_register_custom_op
+else:
+    from vllm.utils.torch_utils import direct_register_custom_op
+
 
 class AscendAttentionBackend(AttentionBackend):
     accept_output_buffer: bool = True

vllm_ascend/models/deepseek_v3_2.py

@@ -67,7 +67,6 @@
 from vllm_ascend.models.layers.sfa import AscendSFAModules, Indexer
 from vllm_ascend.ops.common_fused_moe import AscendFusedMoE
 from vllm_ascend.ops.linear import AscendLinearBase
-
 from vllm_ascend.utils import vllm_version_is
 
 if vllm_version_is("0.11.0"):

vllm_ascend/models/layers/mla.py

@@ -28,17 +28,18 @@
 from vllm.config import CacheConfig, get_current_vllm_config
 from vllm.forward_context import ForwardContext, get_forward_context
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.utils import direct_register_custom_op
 
 from vllm_ascend.utils import vllm_version_is
 
 if vllm_version_is("0.11.0"):
     from vllm.attention import Attention
     from vllm.model_executor.layers.mla import \
         MultiHeadLatentAttention as MultiHeadLatentAttentionWrapper
+    from vllm.utils import direct_register_custom_op
 else:
     from vllm.attention.layer import MLAAttention
     from vllm.model_executor.layers.mla import MultiHeadLatentAttentionWrapper
+    from vllm.utils.torch_utils import direct_register_custom_op
 
 
 @dataclass

vllm_ascend/models/layers/sfa.py

@@ -29,17 +29,18 @@
 from vllm.forward_context import ForwardContext, get_forward_context
 from vllm.model_executor.layers.linear import ReplicatedLinear
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.utils import direct_register_custom_op
 
 from vllm_ascend.utils import vllm_version_is
 
 if vllm_version_is("0.11.0"):
     from vllm.attention import Attention
     from vllm.model_executor.layers.mla import \
         MultiHeadLatentAttention as MultiHeadLatentAttentionWrapper
+    from vllm.utils import direct_register_custom_op
 else:
     from vllm.attention.layer import MLAAttention
     from vllm.model_executor.layers.mla import MultiHeadLatentAttentionWrapper
+    from vllm.utils.torch_utils import direct_register_custom_op
 
 
 @dataclass

vllm_ascend/ops/register_custom_ops.py

@@ -8,12 +8,17 @@
                               tensor_model_parallel_all_reduce,
                               tensor_model_parallel_reduce_scatter)
 from vllm.forward_context import get_forward_context
-from vllm.utils import direct_register_custom_op
 
 import vllm_ascend.envs as envs_ascend
 from vllm_ascend.ascend_forward_context import MoECommType
 from vllm_ascend.ops.weight_prefetch import maybe_npu_prefetch
-from vllm_ascend.utils import npu_stream_switch, prefetch_stream
+from vllm_ascend.utils import (npu_stream_switch, prefetch_stream,
+                               vllm_version_is)
+
+if vllm_version_is("0.11.0"):
+    from vllm.utils import direct_register_custom_op
+else:
+    from vllm.utils.torch_utils import direct_register_custom_op
 
 
 def _maybe_chunk_residual_impl(x: torch.Tensor,

vllm_ascend/patch/platform/patch_common/patch_mamba_config.py

@@ -3,9 +3,16 @@
 from vllm.logger import init_logger
 from vllm.model_executor.models import ModelRegistry
 from vllm.model_executor.models.config import MambaModelConfig
-from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE, cdiv
+from vllm.utils import cdiv
 from vllm.v1.kv_cache_interface import FullAttentionSpec, MambaSpec
 
+from vllm_ascend.utils import vllm_version_is
+
+if vllm_version_is("0.11.0"):
+    from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE
+else:
+    from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
+
 
 @classmethod
 def verify_and_update_config(cls, vllm_config) -> None:

vllm_ascend/patch/worker/patch_common/patch_weight_loader.py

@@ -3,7 +3,13 @@
 from vllm.logger import init_logger
 from vllm.model_executor.layers.linear import UnquantizedLinearMethod
 from vllm.model_executor.utils import set_weight_attrs
-from vllm.utils import GiB_bytes
+
+from vllm_ascend.utils import vllm_version_is
+
+if vllm_version_is("0.11.0"):
+    from vllm.utils import GiB_bytes
+else:
+    from vllm.utils.mem_constants import GiB_bytes
 
 logger = init_logger(__name__)