-
Notifications
You must be signed in to change notification settings - Fork 162
/
train.py
executable file
·196 lines (167 loc) · 7.3 KB
/
train.py
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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
#!/usr/bin/env python
import argparse
import ncluster
import os
IMAGE_NAME = 'pytorch.imagenet.source.v7'
INSTANCE_TYPE = 'p3.16xlarge'
NUM_GPUS = 8
ncluster.set_backend('aws')
parser = argparse.ArgumentParser()
parser.add_argument('--name', type=str, default='imagenet',
help="name of the current run, used for machine naming and tensorboard visualization")
parser.add_argument('--machines', type=int, default=16,
help="how many machines to use")
args = parser.parse_args()
# 109:12 to 93.00
# events: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-1
# logs: https://s3.amazonaws.com/yaroslavvb/logs/imagenet1.tar
lr = 1.0
bs = [512, 224, 128] # largest batch size that fits in memory for each image size
bs_scale = [x/bs[0] for x in bs]
one_machine = [
{'ep':0, 'sz':128, 'bs':bs[0], 'trndir':'-sz/160'},
{'ep':(0,7), 'lr':(lr,lr*2)}, # lr warmup is better with --init-bn0
{'ep':(7,13), 'lr':(lr*2,lr/4)}, # trying one cycle
{'ep':13, 'sz':224, 'bs':bs[1], 'trndir':'-sz/352', 'min_scale':0.087},
{'ep':(13,22),'lr':(lr*bs_scale[1],lr/10*bs_scale[1])},
{'ep':(22,25),'lr':(lr/10*bs_scale[1],lr/100*bs_scale[1])},
{'ep':25, 'sz':288, 'bs':bs[2], 'min_scale':0.5, 'rect_val':True},
{'ep':(25,28),'lr':(lr/100*bs_scale[2],lr/1000*bs_scale[2])}
]
# 29:44 to 93.05
# events: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-4
# logs: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-4.tar
lr = 0.50 * 4 # 4 = num tasks
bs = [256, 224, 128] # largest batch size that fits in memory for each image size
bs_scale = [x/bs[0] for x in bs] # scale learning rate to batch size
four_machines = [
{'ep':0, 'sz':128, 'bs':bs[0], 'trndir':'-sz/160'}, # bs = 256 * 4 * 8 = 8192
{'ep':(0,6), 'lr':(lr,lr*2)},
{'ep':6, 'sz':128, 'bs':bs[0]*2, 'keep_dl':True},
{'ep':6, 'lr':lr*2},
{'ep':(11,13), 'lr':(lr*2,lr)}, # trying one cycle
{'ep':13, 'sz':224, 'bs':bs[1], 'trndir': '-sz/352', 'min_scale': 0.087},
{'ep':13, 'lr':lr*bs_scale[1]},
{'ep':(16,24),'lr':(lr*bs_scale[1],lr/10*bs_scale[1])},
{'ep':(24,28),'lr':(lr/10*bs_scale[1],lr/100*bs_scale[1])},
{'ep':28, 'sz':288, 'bs':bs[2], 'min_scale':0.5, 'rect_val':True},
{'ep':(28,30),'lr':(lr/100*bs_scale[2],lr/1000*bs_scale[2])}
]
# 19:04 to 93.0
# events: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-16.02.8
# logs: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-8.tar
lr = 0.24 * 8
scale_224 = 224/128
eight_machines = [
{'ep':0, 'sz':128, 'bs':128, 'trndir':'-sz/160'},
{'ep':(0,6), 'lr':(lr,lr*2)},
{'ep':6, 'bs':256, 'keep_dl':True,
'lr':lr*2},
{'ep':(11,14),'lr':(lr*2,lr)}, # trying one cycle
{'ep':14, 'sz':224, 'bs':128, 'trndir':'-sz/352', 'min_scale':0.087,
'lr':lr},
{'ep':17, 'bs':224, 'keep_dl':True},
{'ep':(17,23),'lr':(lr,lr/10*scale_224)},
{'ep':(23,29),'lr':(lr/10*scale_224,lr/100*scale_224)},
{'ep':29, 'sz':288, 'bs':128, 'min_scale':0.5, 'rect_val':True},
{'ep':(29,36),'lr':(lr/100,lr/1000)}
]
# 16:08 to 93.04 (after prewarming)
# events: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-16.02.thu16
# logs: https://s3.amazonaws.com/yaroslavvb/logs/imagenet-16.cmd.tar
lr = 0.235 * 8 #
bs = 64
sixteen_machines = [
{'ep':0, 'sz':128, 'bs':64, 'trndir':'-sz/160'},
{'ep':(0,6), 'lr':(lr,lr*2)},
{'ep':6, 'bs':128, 'keep_dl':True},
{'ep':6, 'lr':lr*2},
{'ep':16, 'sz':224,'bs':64}, # todo: increase this bs
{'ep':16, 'lr':lr},
{'ep':19, 'bs':192, 'keep_dl':True},
{'ep':19, 'lr':2*lr/(10/1.5)},
{'ep':31, 'lr':2*lr/(100/1.5)},
{'ep':37, 'sz':288, 'bs':128, 'min_scale':0.5, 'rect_val':True},
{'ep':37, 'lr':2*lr/100},
{'ep':(38,50),'lr':2*lr/1000}
]
schedules = {1: one_machine,
4: four_machines,
8: eight_machines,
16: sixteen_machines}
# routines to build NCCL ring orders
def get_nccl_params(num_tasks, num_gpus):
if num_tasks <= 1:
return 'NCCL_DEBUG=VERSION'
nccl_rings = get_nccl_rings(num_tasks, num_gpus)
return f'NCCL_RINGS="{nccl_rings}" NCCL_SINGLE_RING_THRESHOLD=10 NCCL_DEBUG=VERSION'
# return 'NCCL_MIN_NRINGS=2 NCCL_SINGLE_RING_THRESHOLD=10 NCCL_DEBUG=VERSION'
def get_nccl_rings(num_tasks, num_gpus):
ring = build_ring_order(range(num_tasks), range(num_gpus))
ring_rev = build_ring_order(reversed(range(num_tasks)),
reversed(range(num_gpus)))
rotated_gpu_order = [3, 2, 1, 0, 7, 6, 5, 4]
skip_gpu_order = get_skip_order(num_gpus)
if (num_tasks >= 4) and (num_gpus == 8):
assert ((num_tasks % 4) == 0)
skip_machine_order = get_skip_order(num_tasks)
ring_skip = build_ring_order(skip_machine_order, rotated_gpu_order)
ring_skip_rev = build_ring_order(reversed(skip_machine_order),
skip_gpu_order)
rings_arr = [ring, ring_rev, ring_skip, ring_skip_rev]
# rings_arr = [ring, ring_rev, ring_skip]
else:
rings_arr = [ring, ring_rev]
return ' | '.join(rings_arr)
def build_ring_order(machine_order, gpu_order):
gpu_order = list(gpu_order)
machine_order = list(machine_order)
ngpus = len(gpu_order)
r_order = [(x * ngpus) + y for x in machine_order for y in gpu_order]
return ' '.join(map(str, r_order))
def get_skip_order(size):
if size == 4:
return [0, 2, 1, 3]
skip_step = 5 if size == 16 else 3
# step size of 3 yields - [0,3,6,1,4,7,2,5]
return [(i * skip_step) % size for i in range(size)]
def format_params(arg):
if isinstance(arg, list) or isinstance(arg, dict):
return '\"' + str(arg) + '\"'
else:
return str(arg)
def main():
supported_regions = ['us-west-2', 'us-east-1', 'us-east-2']
assert ncluster.get_region() in supported_regions, f"required AMI {IMAGE_NAME} has only been made available in regions {supported_regions}, but your current region is {ncluster.get_region()}"
assert args.machines in schedules, f"{args.machines} not supported, only support {schedules.keys()}"
os.environ['NCLUSTER_AWS_FAST_ROOTDISK'] = '1' # use io2 disk on AWS
job = ncluster.make_job(name=args.name,
run_name=f"{args.name}-{args.machines}",
num_tasks=args.machines,
image_name=IMAGE_NAME,
instance_type=INSTANCE_TYPE,
install_script=open('setup.sh').read())
job.upload('training')
job.run(f'source activate pytorch_source')
nccl_params = get_nccl_params(args.machines, NUM_GPUS)
# Training script args
default_params = [
'~/data/imagenet',
'--fp16',
'--logdir', job.logdir,
'--distributed',
'--init-bn0',
'--no-bn-wd',
]
params = ['--phases', schedules[args.machines]]
training_params = default_params + params
training_params = ' '.join(map(format_params, training_params))
# TODO: simplify args processing, or give link to actual commands run
for i, task in enumerate(job.tasks):
dist_params = f'--nproc_per_node=8 --nnodes={args.machines} --node_rank={i} --master_addr={job.tasks[0].ip} --master_port={6006}'
cmd = f'{nccl_params} python -m torch.distributed.launch {dist_params} training/train_imagenet_nv.py {training_params}'
task.run(f'echo {cmd} > {job.logdir}/task-{i}.cmd') # save command-line
task.run(cmd, non_blocking=True)
print(f"Logging to {job.logdir}")
if __name__ == '__main__':
main()