Russell Bryant
e489ad7a21
- **Add SPDX license headers to python source files** - **Check for SPDX headers using pre-commit** commit 9d7ef44c3cfb72ca4c32e1c677d99259d10d4745 Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:18:24 2025 -0500 Add SPDX license headers to python source files This commit adds SPDX license headers to python source files as recommended to the project by the Linux Foundation. These headers provide a concise way that is both human and machine readable for communicating license information for each source file. It helps avoid any ambiguity about the license of the code and can also be easily used by tools to help manage license compliance. The Linux Foundation runs license scans against the codebase to help ensure we are in compliance with the licenses of the code we use, including dependencies. Having these headers in place helps that tool do its job. More information can be found on the SPDX site: - https://spdx.dev/learn/handling-license-info/ Signed-off-by: Russell Bryant <rbryant@redhat.com> commit 5a1cf1cb3b80759131c73f6a9dddebccac039dea Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:36:32 2025 -0500 Check for SPDX headers using pre-commit Signed-off-by: Russell Bryant <rbryant@redhat.com> --------- Signed-off-by: Russell Bryant <rbryant@redhat.com>
77 lines
2.6 KiB
Python
77 lines
2.6 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
"""Utilities for selecting and loading models."""
|
|
import contextlib
|
|
from dataclasses import dataclass, field
|
|
from typing import Dict, List, Tuple, Type
|
|
|
|
import torch
|
|
from torch import nn
|
|
|
|
from vllm.config import ModelConfig
|
|
from vllm.model_executor.models import ModelRegistry
|
|
from vllm.model_executor.models.adapters import (as_classification_model,
|
|
as_embedding_model,
|
|
as_reward_model)
|
|
|
|
|
|
@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 get_model_architecture(
|
|
model_config: ModelConfig) -> Tuple[Type[nn.Module], str]:
|
|
architectures = getattr(model_config.hf_config, "architectures", [])
|
|
|
|
# Special handling for quantized Mixtral.
|
|
# FIXME(woosuk): This is a temporary hack.
|
|
mixtral_supported = [
|
|
"fp8", "compressed-tensors", "gptq_marlin", "awq_marlin"
|
|
]
|
|
|
|
if (model_config.quantization is not None
|
|
and model_config.quantization not in mixtral_supported
|
|
and "MixtralForCausalLM" in architectures):
|
|
architectures = ["QuantMixtralForCausalLM"]
|
|
|
|
model_cls, arch = ModelRegistry.resolve_model_cls(architectures)
|
|
if model_config.task == "embed":
|
|
model_cls = as_embedding_model(model_cls)
|
|
elif model_config.task == "classify":
|
|
model_cls = as_classification_model(model_cls)
|
|
elif model_config.task == "reward":
|
|
model_cls = as_reward_model(model_cls)
|
|
|
|
return model_cls, arch
|
|
|
|
|
|
def get_architecture_class_name(model_config: ModelConfig) -> str:
|
|
return get_model_architecture(model_config)[1]
|
|
|
|
|
|
@dataclass
|
|
class ParamMapping:
|
|
"""
|
|
A class to handle parameter mapping for model weight loading.
|
|
It creates a bidirectional mapping between packed parameters and their
|
|
constituent parts.
|
|
"""
|
|
packed_mapping: Dict[str, List[str]]
|
|
inverse_packed_mapping: Dict[str, Tuple[str,
|
|
int]] = field(default_factory=dict)
|
|
|
|
def __post_init__(self):
|
|
for packed_name, sub_params in self.packed_mapping.items():
|
|
# Skip self-contained cases (e.g., {"W_pack": ["W_pack"]})
|
|
if len(sub_params) == 1 and sub_params[0] == packed_name:
|
|
continue
|
|
for index, param_name in enumerate(sub_params):
|
|
self.inverse_packed_mapping[param_name] = (
|
|
packed_name,
|
|
index,
|
|
)
|