Remove unused imports (#26972)

Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>

vllm/model_executor/layers/mamba/linear_attn.py

@@ -10,7 +10,6 @@ if TYPE_CHECKING:
 from typing import TYPE_CHECKING
 
 import torch
-import torch.distributed
 import torch.nn.functional as F
 from einops import rearrange
 from torch import nn
@@ -41,9 +40,6 @@ from vllm.v1.attention.backends.linear_attn import LinearAttentionMetadata
 if TYPE_CHECKING:
     from vllm.attention.backends.abstract import AttentionBackend
 
-import torch
-import torch.distributed
-
 
 class MiniMaxText01RMSNormTP(CustomOp):
     name = "MiniMaxText01RMSNormTP"

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py

@@ -848,7 +848,6 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod):
         # Property to determine if AITER is used
         if self.rocm_aiter_moe_enabled:
             from vllm.model_executor.layers.fused_moe.rocm_aiter_fused_moe import (  # noqa E501
-                rocm_aiter_fused_experts,
                 shuffle_weights,
             )
 

vllm/model_executor/models/gemma3n_mm.py

@@ -58,7 +58,6 @@ from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsTransc
 from .utils import (
     AutoWeightsLoader,
     WeightsMapper,
-    flatten_bn,
     init_vllm_registered_model,
     maybe_prefix,
 )

vllm/model_executor/models/minimax_text_01.py

@@ -11,7 +11,6 @@ if TYPE_CHECKING:
 
 import regex as re
 import torch
-import torch.distributed
 from torch import nn
 from transformers import MiniMaxConfig
 

vllm/multimodal/utils.py

@@ -31,13 +31,11 @@ if TYPE_CHECKING:
     from .inputs import (
         BatchedTensorInputs,
         MultiModalKwargsItem,
-        MultiModalKwargsItems,
         MultiModalPlaceholderDict,
     )
 else:
     BatchedTensorInputs = Any
     MultiModalKwargsItem = Any
-    MultiModalKwargsItems = Any
     MultiModalPlaceholderDict = Any
 
 global_thread_pool = ThreadPoolExecutor(

vllm/transformers_utils/tokenizer.py

@@ -21,11 +21,9 @@ from vllm.transformers_utils.utils import check_gguf_file
 
 if TYPE_CHECKING:
     from vllm.config import ModelConfig
-    from vllm.lora.request import LoRARequest
     from vllm.transformers_utils.tokenizer_base import TokenizerBase
 else:
     ModelConfig = Any
-    LoRARequest = Any
     TokenizerBase = Any
 
 logger = init_logger(__name__)

vllm/utils/__init__.py

@@ -900,7 +900,6 @@ def _cuda_device_count_stateless(cuda_visible_devices: str | None = None) -> int
     # c1cd946818442aca8c7f812b16d187ce1586c3bc/
     # torch/cuda/__init__.py#L831C1-L831C17
     import torch.cuda
-    import torch.version
 
     from vllm.platforms import current_platform
 

vllm/v1/worker/tpu_worker.py

@@ -7,7 +7,6 @@ from collections.abc import Callable
 from typing import Any, TypeVar
 
 import torch
-import torch.distributed
 import torch.nn as nn
 
 import vllm.envs as envs