remove unused code

vllm/attention/backends/blocksparse_attn.py

@@ -13,7 +13,6 @@ from vllm.attention.backends.utils import (CommonAttentionState,
 from vllm.attention.ops.blocksparse_attention.interface import (
     LocalStridedBlockSparseAttn, get_head_sliding_step)
 from vllm.attention.ops.paged_attn import PagedAttention
-from vllm import _custom_ops as ops
 from vllm.distributed import (get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
 

vllm/attention/backends/rocm_flash_attn.py

@@ -148,7 +148,6 @@ class ROCmFlashAttentionMetadata(AttentionMetadata, PagedAttentionMetadata):
     cross_slot_mapping: Optional[torch.Tensor] = None
     cross_block_tables: Optional[torch.Tensor] = None
 
-
     @property
     def prefill_metadata(self) -> Optional["ROCmFlashAttentionMetadata"]:
         if self.num_prefills == 0:
@@ -524,10 +523,10 @@ class ROCmFlashAttentionImpl(AttentionImpl):
             logger.debug("Using Triton FA in ROCmBackend")
             if self.sliding_window != (-1, -1):
                 logger.warning("ROCm Triton FA does not currently support "
-                            "sliding window attention. If using half "
-                            "precision, please try using the ROCm CK "
-                            "FA backend instead by setting the env var "
-                            "`VLLM_USE_TRITON_FLASH_ATTN=0`")
+                               "sliding window attention. If using half "
+                               "precision, please try using the ROCm CK "
+                               "FA backend instead by setting the env var "
+                               "`VLLM_USE_TRITON_FLASH_ATTN=0`")
         else:
             # if not using triton, navi3x/navi21/navi10 do not use flash-attn
             # either
@@ -723,7 +722,6 @@ class ROCmFlashAttentionImpl(AttentionImpl):
                         attn_masks[0][None]
                         if attn_masks is not None else None,
                     )
-                
                 elif self.use_naive_attn:
                     if self.num_kv_heads != self.num_heads:
                         # Interleave for MQA workaround.

vllm/config.py

@@ -2305,7 +2305,6 @@ class SpeculativeConfig:
                 f"other value than 1 or target model tensor_parallel_size")
         return speculative_draft_tensor_parallel_size
 
-
     @staticmethod
     def create_draft_parallel_config(
         target_parallel_config: ParallelConfig,

vllm/distributed/device_communicators/custom_all_reduce.py

@@ -19,7 +19,6 @@ from vllm.utils import cuda_device_count_stateless
 try:
     ops.meta_size()
     custom_ar = True
-
 except Exception:
     # For CPUs
     custom_ar = False
@@ -130,7 +129,6 @@ class CustomAllreduce:
         # test nvlink first, this will filter out most of the cases
         # where custom allreduce is not supported
         # this checks hardware and driver support for NVLink
-        
         assert current_platform.is_cuda_alike()
         fully_connected = current_platform.is_fully_connected(
             physical_device_ids)

vllm/lora/models.py

@@ -341,6 +341,7 @@ class LoRAModelManager(AdapterModelManager):
         # Used for long context lora.
         self.scaling_factor_to_offset: Dict[float, int] = {}
         super().__init__(model)
+        
         self.supported_lora_modules = get_supported_lora_modules(self.model)
         assert self.supported_lora_modules, "No supported LoRA modules found in"
         f"{self.model.__class__.__name__}."

vllm/multimodal/utils.py

@@ -9,7 +9,6 @@ import numpy as np
 import numpy.typing as npt
 import torch
 from PIL import Image
-import os
 
 import vllm.envs as envs
 from vllm.connections import HTTPConnection, global_http_connection
@@ -87,7 +86,6 @@ class MediaConnector:
 
         return media_io.load_file(filepath)
 
-
     def load_from_url(
         self,
         url: str,

vllm/platforms/cpu.py

@@ -31,10 +31,6 @@ class CpuPlatform(Platform):
     @classmethod
     def get_device_name(cls, device_id: int = 0) -> str:
         return "cpu"
-    
-    @classmethod
-    def get_device_total_memory(cls, device_id: int = 0) -> int:
-        return psutil.virtual_memory().total
 
     @classmethod
     def get_attn_backend_cls(cls, selected_backend: _Backend, head_size: int,

vllm/platforms/interface.py

@@ -203,11 +203,6 @@ class Platform:
         """
         raise NotImplementedError
 
-    @classmethod
-    def get_device_total_memory(cls, device_id: int = 0) -> int:
-        """Get the total memory of a device in bytes."""
-        raise NotImplementedError
-
     @classmethod
     def inference_mode(cls):
         """A device-specific wrapper of `torch.inference_mode`.

vllm/transformers_utils/configs/qwen2vl.py

-# coding=utf-8
-# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team.
-# All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""Qwen2VL model configuration"""
-
-import os
-from typing import Union
-
-from transformers import PretrainedConfig
-
-
-class Qwen2VLVisionConfig(PretrainedConfig):
-    model_type = "qwen2_vl"
-
-    def __init__(
-        self,
-        depth=32,
-        embed_dim=1280,
-        hidden_size=3584,
-        hidden_act="quick_gelu",
-        mlp_ratio=4,
-        num_heads=16,
-        in_channels=3,
-        patch_size=14,
-        spatial_merge_size=2,
-        temporal_patch_size=2,
-        **kwargs,
-    ):
-        super().__init__(**kwargs)
-
-        self.depth = depth
-        self.embed_dim = embed_dim
-        self.hidden_size = hidden_size
-        self.hidden_act = hidden_act
-        self.mlp_ratio = mlp_ratio
-        self.num_heads = num_heads
-        self.in_channels = in_channels
-        self.patch_size = patch_size
-        self.spatial_merge_size = spatial_merge_size
-        self.temporal_patch_size = temporal_patch_size
-
-    @classmethod
-    def from_pretrained(cls, pretrained_model_name_or_path: Union[str,
-                                                                  os.PathLike],
-                        **kwargs) -> "PretrainedConfig":
-        cls._set_token_in_kwargs(kwargs)
-
-        config_dict, kwargs = cls.get_config_dict(
-            pretrained_model_name_or_path, **kwargs)
-
-        if config_dict.get("model_type") == "qwen2_vl":
-            config_dict = config_dict["vision_config"]
-
-        return cls.from_dict(config_dict, **kwargs)
-
-
-class Qwen2VLConfig(PretrainedConfig):
-
-    def __init__(
-        self,
-        vocab_size=152064,
-        hidden_size=8192,
-        intermediate_size=29568,
-        num_hidden_layers=80,
-        num_attention_heads=64,
-        num_key_value_heads=8,
-        hidden_act="silu",
-        max_position_embeddings=32768,
-        initializer_range=0.02,
-        rms_norm_eps=1e-05,
-        use_cache=True,
-        tie_word_embeddings=False,
-        rope_theta=1000000.0,
-        use_sliding_window=False,
-        sliding_window=4096,
-        max_window_layers=80,
-        attention_dropout=0.0,
-        vision_config=None,
-        rope_scaling=None,
-        **kwargs,
-    ):
-        if isinstance(vision_config, dict):
-            self.vision_config = Qwen2VLVisionConfig(**vision_config)
-        elif vision_config is None:
-            self.vision_config = Qwen2VLVisionConfig()
-
-        self.vocab_size = vocab_size
-        self.max_position_embeddings = max_position_embeddings
-        self.hidden_size = hidden_size
-        self.intermediate_size = intermediate_size
-        self.num_hidden_layers = num_hidden_layers
-        self.num_attention_heads = num_attention_heads
-        self.use_sliding_window = use_sliding_window
-        self.sliding_window = sliding_window
-        self.max_window_layers = max_window_layers
-
-        # for backward compatibility
-        if num_key_value_heads is None:
-            num_key_value_heads = num_attention_heads
-
-        self.num_key_value_heads = num_key_value_heads
-        self.hidden_act = hidden_act
-        self.initializer_range = initializer_range
-        self.rms_norm_eps = rms_norm_eps
-        self.use_cache = use_cache
-        self.rope_theta = rope_theta
-        self.attention_dropout = attention_dropout
-        self.rope_scaling = rope_scaling
-
-        # NOTE: the following section from original transformers config
-        # for Qwen2-VL is commented out to address rope config loading issue
-        #
-        # if self.rope_scaling is not None and "type" in self.rope_scaling:
-        #     if self.rope_scaling["type"] == "mrope":
-        #         self.rope_scaling["type"] = "default"
-        #     self.rope_scaling["rope_type"] = self.rope_scaling["type"]
-        # rope_config_validation(self)
-
-        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
\ No newline at end of file

vllm/worker/model_runner.py

@@ -352,6 +352,7 @@ class ModelInputForGPUBuilder(ModelRunnerInputBuilderBase[ModelInputForGPU]):
                             prompt_adapter_prompt_mapping
                     else:
                         self.prompt_adapter_prompt_mapping.clear()
+                        
             else:
                 self.input_tokens = input_tokens or []
                 self.input_positions = input_positions or []