[Mypy] Fix mypy typing for Transformers models (#54320)

Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Taneem Ibrahim
2026-09-17 12:11:31 -04:00
committed by GitHub
co-authored by Harry Mellor
parent b78fbab4be
commit e960ead0aa
16 changed files with 224 additions and 145 deletions
@@ -18,11 +18,10 @@
from collections.abc import Callable, Iterator
from contextlib import contextmanager
from functools import lru_cache
from itertools import chain
from operator import attrgetter
from pathlib import Path
from typing import TYPE_CHECKING, Literal, TypeVar
from typing import TYPE_CHECKING, Literal, TypeAlias, TypeVar
import torch
from torch import nn
@@ -70,7 +69,7 @@ def init_on_device_without_buffers(device: torch.device):
module._parameters[name].to(device), **kwargs
)
tensor_constructors_to_patch = {}
tensor_constructors_to_patch: dict[str, Callable] = {}
def patch_tensor_constructor(fn):
def wrapper(*args, **kwargs):
@@ -147,8 +146,8 @@ def replace_linear_class(
)
TorchConv = nn.Conv2d | nn.Conv3d
VllmConv = Conv2dLayer | Conv3dLayer
TorchConv: TypeAlias = nn.Conv2d | nn.Conv3d
VllmConv: TypeAlias = Conv2dLayer | Conv3dLayer
def replace_conv_class(conv: TorchConv) -> VllmConv | TorchConv:
@@ -212,7 +211,9 @@ class _VocabParallelEmbeddingBase(VocabParallelEmbedding, _UninitializedEmbeddin
`super().forward(...)` in an `nn.Embedding` subclass reaches vLLM's embedding."""
@lru_cache
_rebased_embedding_classes: dict[type[nn.Embedding], type[VocabParallelEmbedding]] = {}
def _rebase_on_vocab_parallel(cls: type[nn.Embedding]) -> type[VocabParallelEmbedding]:
"""Subclass `cls` so that `VocabParallelEmbedding` supersedes its `nn.Embedding`.
@@ -224,7 +225,10 @@ def _rebase_on_vocab_parallel(cls: type[nn.Embedding]) -> type[VocabParallelEmbe
The new class, to assign to `__class__` of an instance of `cls`.
"""
return type(cls.__name__, (cls, _VocabParallelEmbeddingBase), {})
if cls not in _rebased_embedding_classes:
rebased_cls = type(cls.__name__, (cls, _VocabParallelEmbeddingBase), {})
_rebased_embedding_classes[cls] = rebased_cls
return _rebased_embedding_classes[cls]
def replace_embedding_class(
@@ -278,10 +282,9 @@ def recursive_replace_linear(
qual_name = maybe_prefix(prefix, child_name)
# Replace modules as needed
if isinstance(child_module, nn.Linear):
style = "replicate"
new_module = replace_linear_class(
child_module,
style,
"replicate",
quant_config,
prefix=qual_name,
)