1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
|
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
from bitsandbytes.optim.optimizer import Optimizer2State
class AdamW(Optimizer2State):
def __init__(
self,
params,
lr=1e-3,
betas=(0.9, 0.999),
eps=1e-8,
weight_decay=1e-2,
amsgrad=False,
optim_bits=32,
args=None,
min_8bit_size=4096,
percentile_clipping=100,
block_wise=True,
):
super(AdamW, self).__init__(
"adam",
params,
lr,
betas,
eps,
weight_decay,
optim_bits,
args,
min_8bit_size,
percentile_clipping,
block_wise,
)
class AdamW8bit(Optimizer2State):
def __init__(
self,
params,
lr=1e-3,
betas=(0.9, 0.999),
eps=1e-8,
weight_decay=1e-2,
amsgrad=False,
args=None,
min_8bit_size=4096,
percentile_clipping=100,
block_wise=True,
):
super(AdamW8bit, self).__init__(
"adam",
params,
lr,
betas,
eps,
weight_decay,
8,
args,
min_8bit_size,
percentile_clipping,
block_wise,
)
class AdamW32bit(Optimizer2State):
def __init__(
self,
params,
lr=1e-3,
betas=(0.9, 0.999),
eps=1e-8,
weight_decay=1e-2,
amsgrad=False,
args=None,
min_8bit_size=4096,
percentile_clipping=100,
block_wise=True,
):
super(AdamW32bit, self).__init__(
"adam",
params,
lr,
betas,
eps,
weight_decay,
32,
args,
min_8bit_size,
percentile_clipping,
block_wise,
)
|
