github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-05-03 14:08:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

[checkpointio] fix async io ()

Browse Source
This commit is contained in:
flybird11111 2024-12-16 10:36:28 +08:00 committed by GitHub
parent de3d371f65
commit e994c64568
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 2 additions and 3 deletions
colossalai/checkpoint_io
general_checkpoint_io.pyutils.py

3
colossalai/checkpoint_io/general_checkpoint_io.py
View File

@ -8,8 +8,6 @@ from typing import Optional
import torch.nn as nn import torch.nn as nn
from torch.optim import Optimizer from torch.optim import Optimizer
from colossalai.utils.safetensors import move_and_save
from .checkpoint_io_base import CheckpointIO from .checkpoint_io_base import CheckpointIO
from .index_file import CheckpointIndexFile from .index_file import CheckpointIndexFile
from .utils import ( from .utils import (
@ -54,6 +52,7 @@ class GeneralCheckpointIO(CheckpointIO):
pass pass
if use_async: if use_async:
from colossalai.utils.safetensors import move_and_save
if id(model) not in self.pinned_state_dicts: if id(model) not in self.pinned_state_dicts:
self.pinned_state_dicts[id(model)] = create_pinned_state_dict(state_dict) self.pinned_state_dicts[id(model)] = create_pinned_state_dict(state_dict)

2
colossalai/checkpoint_io/utils.py
View File

@ -19,7 +19,6 @@ from colossalai.tensor.d_tensor import (
to_global, to_global,
to_global_for_customized_distributed_tensor, to_global_for_customized_distributed_tensor,
) )
from colossalai.utils.safetensors import move_and_save
SAFE_WEIGHTS_NAME = "model.safetensors" SAFE_WEIGHTS_NAME = "model.safetensors"
WEIGHTS_NAME = "pytorch_model.bin" WEIGHTS_NAME = "pytorch_model.bin"
@ -289,6 +288,7 @@ def async_save_state_dict_shards(
Returns: Returns:
int: the total size of shards int: the total size of shards
""" """
from colossalai.utils.safetensors import move_and_save
total_size = 0 total_size = 0
shard_filenames = [] shard_filenames = []