-
Notifications
You must be signed in to change notification settings - Fork 3
/
train_mul_g9_tfgrid_progressive_pitch.py
255 lines (211 loc) · 10 KB
/
train_mul_g9_tfgrid_progressive_pitch.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
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
'''
multi gpu version
'''
import os
os.environ["CUDA_VISIBLE_DEVICES"]="3,4,5,6,7,2,0,1"
import toml
import torch
import torch.nn as nn
import argparse
import torch.distributed as dist
from collections import OrderedDict
from trainer_mul_g9_tfgrid_progressive_pitch import Trainer
from SISO_TFGridNet_progressive import TFGridNet0, TFGridNetPret, TFGridNetMerge2, TFGridNetPE, TFGridNetPEPret
from Dataloader_d_g8_progressive import Dataset
from binaural_loss import loss_MTFAA_t,loss_wavmag, loss_MTFAA_fonly, loss_MTFAA_fonly_asym, loss_MTFAA_t_progressive
from pitch_estimation.pitch_estimator_model import pitch_estimator
from pitch_estimation.comb_filter import comb_filter
def run(rank,config,args):
os.environ['MASTER_ADDR'] = 'localhost'
os.environ['MASTER_PORT'] = '12368'
dist.init_process_group("nccl", rank=rank, world_size=args.world_size)
torch.cuda.set_device(rank)
dist.barrier()
args.rank = rank
args.device = torch.device(rank)
train_dataset = Dataset(**config['train_dataset'])
train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
train_dataloader = torch.utils.data.DataLoader(dataset=train_dataset, sampler=train_sampler,
**config['train_dataloader'])
validation_dataset = Dataset(**config['validation_dataset'])
validation_sampler = torch.utils.data.distributed.DistributedSampler(validation_dataset)
validation_dataloader = torch.utils.data.DataLoader(dataset=validation_dataset, sampler=validation_sampler,
**config['validation_dataloader'])
model_pl = TFGridNetPE()
model_pret = TFGridNetPEPret()
model_pitch = pitch_estimator()
model_comb = comb_filter()
model = TFGridNetMerge2()
# Clone model parameters and buffers
for name, param in model_comb.named_parameters():
cloned_param = param.data.clone().detach().to(args.device)
# exec("model_comb."+name+"=nn.Parameter(cloned_param)")
model_comb.name = nn.Parameter(cloned_param)
if param._grad is not None:
#param._grad.data = param._grad.data.clone().detach().to(args.device)
model_comb.name._grad = param._grad.clone().detach().to(args.device)
for name, buffer in model_comb.named_buffers():
cloned_buffer = buffer.data.clone().detach().to(args.device)
model_comb.name = cloned_buffer
# 模型初始化
ckpoints = '/data/ssd1/tong.lei/exp_FTJNF/SISO_TFgrid_PE_Progressive_allnoise2_2024-07-12-23h19m/checkpoints/model_0037.tar'
checkpoint = torch.load(ckpoints, map_location='cpu')
new_state_dict = OrderedDict()
for k, v in checkpoint['model'].items():
name = k[7:] # remove `module.`
new_state_dict[name] = v #新字典的key值对应的value为一一对应的值。
model_pl.load_state_dict(new_state_dict)#(checkpoint['model'])# g6 dont change
model_pret.conv.load_state_dict(model_pl.conv.state_dict())
model_pret.pe.load_state_dict(model_pl.pe.state_dict())
for x in range(model_pret.n_layers):
model_pret.blocks[x].load_state_dict(model_pl.blocks[x].state_dict())
model_pret.deconvs[x].load_state_dict(model_pl.deconvs[x].state_dict())
checkpoint_path = '/data/ssd1/tong.lei/exp_FTJNF/Pitch_estimator/chkpt/epoch_12_trainloss_0.061130986_validloss_0.061626785.pth'
checkpoint = torch.load(checkpoint_path, map_location='cpu')
new_state_dict = OrderedDict()
for k, v in checkpoint['state_dict'].items():
name = k[7:] # remove `module.`
new_state_dict[name] = v
model_pitch.load_state_dict(new_state_dict)
model.block.load_state_dict(model_pl.blocks[-1].state_dict())
model.deconv.load_state_dict(model_pl.deconvs[-1].state_dict())
model_pret.to(args.device)
model_pitch.to(args.device)
model_comb.to(args.device)
model_comb.window = model_comb.window.clone().detach().to(args.device)
# model_pret.eval()
# model_pitch.eval()
# model_comb.eval()
model.to(args.device)
# 转为DDP模型
model_pret = torch.nn.parallel.DistributedDataParallel(model_pret, device_ids=[rank])
model_pitch = torch.nn.parallel.DistributedDataParallel(model_pitch, device_ids=[rank])
model_comb = torch.nn.parallel.DistributedDataParallel(model_comb, device_ids=[rank])
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[rank])
#optimizer = torch.optim.Adam(params=model.parameters(), lr=config['optimizer']['lr'])
optimizer = NoamOpt(model_size=config['network_config']['model_size'], factor=0.8, warmup=3000,
optimizer=torch.optim.Adam(model.parameters(), lr=0, betas=(0.9, 0.98), eps=1e-9))#一般warmup=step左右
if config['loss']['loss_func'] == 'MTFAA':
loss = loss_MTFAA_t()
elif config['loss']['loss_func'] == 'wavmag':
loss = loss_wavmag()
else:
raise NotImplementedError
trainer = Trainer(config=config, model=[model_pret, model_pitch, model_comb, model],optimizer=optimizer, loss_func=loss,
train_dataloader=train_dataloader, validation_dataloader=validation_dataloader,
train_sampler=train_sampler,args=args)
trainer.train()
dist.destroy_process_group()
class NoamOpt:
"Optim wrapper that implements rate."
def __init__(self, model_size, factor, warmup, optimizer):
self.optimizer = optimizer
self._step = 1
self.warmup = warmup
self.factor = factor
self.model_size = model_size
self._rate = 0
def step(self):
"Update parameters and rate"
self._step += 1
if self._step == 2000000000:
self._step = 1
#self.factor = self.factor * 0.9
rate = self.rate()
for p in self.optimizer.param_groups:
p['lr'] = rate
self._rate = rate
self.optimizer.step()
def rate(self, step = None):
"Implement `lrate` above"
if step is None:
step = self._step
return self.factor * (self.model_size ** (-0.5) * min(step ** (-0.5), step * self.warmup ** (-1.5)))#step = warmnp时，两个相等
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('-C', '--config', default='den_g9_tfgrid.toml')
args = parser.parse_args()
config = toml.load(args.config)
args.world_size = config['DDP']['world_size']
torch.multiprocessing.spawn(
run, args=(config, args,), nprocs=args.world_size, join=True)
'''
multi gpu version
import os
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
import toml
import torch
import argparse
import torch.distributed as dist
from trainer_mul_g9_tfgrid_1 import Trainer
from MISO_TFGridNet import TFGridNet_nostd
from datasets_g9 import dataset_train_t,dataset_testaudio_t
from binaural_loss import loss_MTFAA_t,loss_wavmag
def run(rank,config,args):
os.environ['MASTER_ADDR'] = 'localhost'
os.environ['MASTER_PORT'] = '12345'
dist.init_process_group("nccl", rank=rank, world_size=args.world_size)
torch.cuda.set_device(rank)
dist.barrier()
args.rank = rank
args.device = torch.device(rank)
train_dataset = dataset_train_t(**config['train_dataset'])
train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
train_dataloader = torch.utils.data.DataLoader(dataset=train_dataset, sampler=train_sampler,
**config['train_dataloader'])
validation_dataset = dataset_testaudio_t(**config['validation_dataset'])
validation_sampler = torch.utils.data.distributed.DistributedSampler(validation_dataset)
validation_dataloader = torch.utils.data.DataLoader(dataset=validation_dataset, sampler=validation_sampler,
**config['validation_dataloader'])
model = TFGridNet_nostd()
model.to(args.device)
# 转为DDP模型
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[rank])
#optimizer = torch.optim.Adam(params=model.parameters(), lr=config['optimizer']['lr'])
optimizer = NoamOpt(model_size=config['network_config']['model_size'], factor=1.0, warmup=20000,
optimizer=torch.optim.Adam(model.parameters(), lr=0, betas=(0.9, 0.98), eps=1e-9))#一般warmup=step左右
if config['loss']['loss_func'] == 'MTFAA':
loss = loss_MTFAA_t()
elif config['loss']['loss_func'] == 'wavmag':
loss = loss_wavmag()
else:
raise NotImplementedError
trainer = Trainer(config=config, model=model,optimizer=optimizer, loss_func=loss,
train_dataloader=train_dataloader, validation_dataloader=validation_dataloader,
train_sampler=train_sampler,args=args)
trainer.train()
dist.destroy_process_group()
class NoamOpt:
"Optim wrapper that implements rate."
def __init__(self, model_size, factor, warmup, optimizer):
self.optimizer = optimizer
self._step = 4000
self.warmup = warmup
self.factor = factor
self.model_size = model_size
self._rate = 0
def step(self):
"Update parameters and rate"
self._step += 1
if self._step == 120000:
self._step = 1
#self.factor = self.factor * 0.9
rate = self.rate()
for p in self.optimizer.param_groups:
p['lr'] = rate
self._rate = rate
self.optimizer.step()
def rate(self, step = None):
"Implement `lrate` above"
if step is None:
step = self._step
return self.factor * (self.model_size ** (-0.5) * min(step ** (-0.5), step * self.warmup ** (-1.5)))#step = warmnp时，两个相等
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('-C', '--config', default='den_g9_tfgrid_1.toml')
args = parser.parse_args()
config = toml.load(args.config)
args.world_size = config['DDP']['world_size']
torch.multiprocessing.spawn(
run, args=(config, args,), nprocs=args.world_size, join=True)
'''