-
Notifications
You must be signed in to change notification settings - Fork 1
/
test.sh
131 lines (105 loc) · 8.18 KB
/
test.sh
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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
##########################################################################################################################
# Commands to run experiments on 16-node ring, training CIFAR-10 with ResNet-20, alpha=0.1
## 1. DSGD -- momentum and scaling are set to 0
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.0 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 2. GUT -- momentum is set to 0 and scaling is set to 0.9
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.0 --scaling=0.9 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 3. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.9 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 4. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.06
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.9 --scaling=0.06 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 32-node ring, training CIFAR-10 with VGG-11, alpha=0.01
## 1. DSGD -- momentum and scaling are set to 0
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.01 --epochs=200 --arch=vgg11 --graph=ring --momentum=0.0 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=vgg11 --world_size=32 --skew=0.01 --graph=ring --seed=12
cd ..
## 2. GUT -- momentum is set to 0 and scaling is set to 0.9
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.01 --epochs=200 --arch=vgg11 --graph=ring --momentum=0.0 --scaling=0.9 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=vgg11 --world_size=32 --skew=0.01 --graph=ring --seed=12
cd ..
## 3. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.01 --epochs=200 --arch=vgg11 --graph=ring --momentum=0.9 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=vgg11 --world_size=32 --skew=0.01 --graph=ring --seed=12
cd ..
## 4. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.08
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.01 --epochs=200 --arch=vgg11 --graph=ring --momentum=0.9 --scaling=0.08 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=vgg11 --world_size=32 --skew=0.01 --graph=ring --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 32-node torus, training CIFAR-10 with ResNet-20, alpha=0.1
## 1. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.1 --epochs=200 --arch=resnet --graph=torus --neighbors=4 --momentum=0.9 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=32 --skew=0.1 --graph=torus --seed=12
cd ..
## 2. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.05
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.1 --epochs=200 --arch=resnet --graph=torus --neighbors=4 --momentum=0.9 --scaling=0.05 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=32 --skew=0.1 --graph=torus --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 32-node dyck, training CIFAR-10 with ResNet-20, alpha=0.1
## 1. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.1 --epochs=200 --arch=resnet --graph=dyck --neighbors=3 --momentum=0.9 --scaling=0.0 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=32 --skew=0.1 --graph=dyck --seed=12
cd ..
## 2. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.05
python trainer.py --lr=0.1 --batch-size=1024 --world_size=32 --skew=0.1 --epochs=200 --arch=resnet --graph=dyck --neighbors=3 --momentum=0.9 --scaling=0.05 --devices=4 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=32 --skew=0.1 --graph=dyck --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 16-node ring, training CIFAR-100 with ResNet-20, alpha=0.1
## 1. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.9 --scaling=0.0 --devices=4 --dataset=cifar100 --classes=100 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 2. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.005
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=200 --arch=resnet --graph=ring --momentum=0.9 --scaling=0.005 --devices=4 --dataset=cifar100 --classes=100 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=resnet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 16-node ring, training Fashion-MNIST with LeNet-5, alpha=0.1
## 1. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=100 --arch=lenet5 --graph=ring --momentum=0.9 --scaling=0.0 --devices=4 --dataset=fmnist --classes=10 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=lenet5 --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 2. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.01
python trainer.py --lr=0.1 --batch-size=512 --world_size=16 --skew=0.1 --epochs=100 --arch=lenet5 --graph=ring --momentum=0.9 --scaling=0.01 --devices=4 --dataset=fmnist --classes=10 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.1 --arch=lenet5 --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
##########################################################################################################################
# Commands to run experiments on 16-node ring, training Imagenette with MobileNet-V2, alpha=0.1
## 1. QG-DSGDm -- momentum is set to 0.9 and scaling is set to 0
python trainer.py --lr=0.01 --batch-size=512 --world_size=16 --skew=0.1 --epochs=100 --arch=mobilenet --graph=ring --momentum=0.9 --scaling=0.0 --devices=4 --dataset=imagenette --classes=10 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.01 --arch=mobilenet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
## 2. QG-GUTm -- momentum is set to 0.9 and scaling is set to 0.03
python trainer.py --lr=0.01 --batch-size=512 --world_size=16 --skew=0.1 --epochs=100 --arch=mobilenet --graph=ring --momentum=0.9 --scaling=0.03 --devices=4 --dataset=imagenette --classes=10 --seed=12
cd ./outputs
python dict_to_csv.py --norm=evonorm --lr=0.01 --arch=mobilenet --world_size=16 --skew=0.1 --graph=ring --seed=12
cd ..
##########################################################################################################################