[Chore] Clean up pytorch helper functions in vllm.utils (#26908)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: isotr0py <2037008807@qq.com>
This commit is contained in:
Isotr0py
2025-10-18 09:48:22 -07:00
committed by GitHub
parent 5c2acb270a
commit 6ac5e06f7c
119 changed files with 772 additions and 714 deletions
@@ -11,8 +11,8 @@ from vllm.logger import init_logger
from vllm.model_executor.model_loader.utils import (
initialize_model,
process_weights_after_loading,
set_default_torch_dtype,
)
from vllm.utils.torch_utils import set_default_torch_dtype
logger = init_logger(__name__)
@@ -32,7 +32,7 @@ from vllm.model_executor.layers.linear import (
RowParallelLinear,
)
from vllm.model_executor.model_loader.base_loader import BaseModelLoader
from vllm.model_executor.model_loader.utils import ParamMapping, set_default_torch_dtype
from vllm.model_executor.model_loader.utils import ParamMapping
from vllm.model_executor.model_loader.weight_utils import (
download_safetensors_index_file_from_hf,
download_weights_from_hf,
@@ -48,6 +48,7 @@ from vllm.model_executor.utils import (
set_weight_attrs,
)
from vllm.platforms import current_platform
from vllm.utils.torch_utils import set_default_torch_dtype
logger = init_logger(__name__)
@@ -15,13 +15,13 @@ from vllm.model_executor.model_loader.base_loader import BaseModelLoader
from vllm.model_executor.model_loader.utils import (
initialize_model,
process_weights_after_loading,
set_default_torch_dtype,
)
from vllm.model_executor.model_loader.weight_utils import (
get_gguf_extra_tensor_names,
get_gguf_weight_type_map,
gguf_quant_weights_iterator,
)
from vllm.utils.torch_utils import set_default_torch_dtype
class GGUFModelLoader(BaseModelLoader):
@@ -22,8 +22,8 @@ from vllm.model_executor.model_loader.tensorizer import (
from vllm.model_executor.model_loader.utils import (
get_model_architecture,
initialize_model,
set_default_torch_dtype,
)
from vllm.utils.torch_utils import set_default_torch_dtype
logger = init_logger(__name__)
+1 -1
View File
@@ -14,8 +14,8 @@ from vllm.model_executor.model_loader.default_loader import DefaultModelLoader
from vllm.model_executor.model_loader.utils import (
initialize_model,
process_weights_after_loading,
set_default_torch_dtype,
)
from vllm.utils.torch_utils import set_default_torch_dtype
logger = init_logger(__name__)
-10
View File
@@ -2,7 +2,6 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""Utilities for selecting and loading models."""
import contextlib
import inspect
import warnings
from contextlib import contextmanager
@@ -32,15 +31,6 @@ from vllm.utils import is_pin_memory_available
logger = init_logger(__name__)
@contextlib.contextmanager
def set_default_torch_dtype(dtype: torch.dtype):
"""Sets the default torch dtype to the given dtype."""
old_dtype = torch.get_default_dtype()
torch.set_default_dtype(dtype)
yield
torch.set_default_dtype(old_dtype)
def initialize_model(
vllm_config: VllmConfig,
*,