Speed up docs build (#44635)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2026-06-05 14:51:44 +00:00
committed by GitHub
parent c66b19800b
commit a80af24356
32 changed files with 234 additions and 159 deletions
+10 -9
View File
@@ -3,6 +3,7 @@
from collections.abc import Callable, Hashable
from fractions import Fraction
from typing import Any
from weakref import WeakValueDictionary
import torch
@@ -42,10 +43,9 @@ class BasevLLMParameter(Parameter):
"""
Initialize the BasevLLMParameter
:param data: torch tensor with the parameter data
:param weight_loader: weight loader callable
:returns: a torch.nn.parameter
Args:
data: torch tensor with the parameter data
weight_loader: weight loader callable
"""
# During weight loading, we often do something like:
@@ -445,15 +445,16 @@ class SharedWeightParameter(BasevLLMParameter):
"currently support tensor parallelism"
)
def add_partition(self, index: int, data_key: Hashable, *args, **kwargs):
def add_partition(self, index: int, data_key: Hashable, *args: Any, **kwargs: Any):
"""
Add a partition to the weight parameter. Partitions whose `data_key`
is the same will share tensor data
:param index: index of partition to add
:param data_key: hashable key used to key shared tensors
:param *args: arguments for `torch.empty`
:param **kwargs: keyword arguments for `torch.empty`
Args:
index: index of partition to add
data_key: hashable key used to key shared tensors
*args: arguments for `torch.empty`
**kwargs: keyword arguments for `torch.empty`
"""
# load (shared) tensor using `data_key`
if data_key not in self.tensors_registry: