-
Notifications
You must be signed in to change notification settings - Fork 3
/
train.py
325 lines (274 loc) · 12.9 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
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
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
# *****************************************************************************
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of the NVIDIA CORPORATION nor the
# names of its contributors may be used to endorse or promote products
# derived from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
# ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# *****************************************************************************
# forked from nv-wavenet/pytorch:
# https://github.com/NVIDIA/nv-wavenet/blob/master/pytorch/train.py
#
# Modified January 2018 by Gary Plunkett for use on the Maestro dataset
import argparse
import json
import os
import time
from csv import DictWriter
import numpy as np
import torch
import torch.nn.functional as F
from torch.autograd import Variable
from scipy.io.wavfile import write
import utils
from utils import as_variable, mu_law_encode
from maestro_dataloader import MaestroDataloader
from scheduled_sampling import ScheduledSamplerWithPatience
import debug
from nn.wavenet import Wavenet
from nn import discretized_mix_logistics as DML
from nn.wavenet_autoencoder import WavenetAutoencoder
from distributed import init_distributed, apply_gradient_allreduce, reduce_tensor
from torch.utils.data.distributed import DistributedSampler
from torch.utils.data import DataLoader
class CrossEntropyLoss(torch.nn.Module):
def __init__(self):
super(CrossEntropyLoss, self).__init__()
self.num_classes = wavenet_config["n_out_channels"]
def forward(self, inputs, targets):
"""
inputs are batch by num_classes by sample
targets are batch by sample
torch CrossEntropyLoss needs
input = batch * samples by num_classes
targets = batch * samples
"""
targets = targets.view(-1)
inputs = inputs.transpose(1, 2)
inputs = inputs.contiguous()
inputs = inputs.view(-1, self.num_classes)
return torch.nn.CrossEntropyLoss()(inputs, targets)
class L2DiversityLoss(torch.nn.Module):
"""
L2 diversity loss as detailed in section 3.2 of "The Challenge of
Realistic Music Generation: Modelling Raw Audio at Scale".
https://arxiv.org/abs/1806.10474
This term encourages the midi autoencoder distribution to be uniform across
dimensions. This output distribution is quantized into a one-hot via argmax,
so making the output distribuion uniform ensures each one-hot vector has an
equal chance of occuring.
"""
def __init__(self):
super(L2DiversityLoss, self).__init__()
def forward(self, q_bar):
"""
Notes on how this works:
q_bar is the continous autoencoder output distribution averaged across batch and time
Each q is normalized so sum(q)=1
Let k be the dimensionality of q
If q_bar is distributed uniformly, q_bar[i] = 1/k
k*q_bar[i] is encouraged to equal 1 using L2 loss
q_bar will be well-estimated, as there are 16,000 instances of q from every
second of training data
"""
k = q_bar.size(0)
loss = torch.sum((k*q_bar - 1) ** 2)
return loss
def load_checkpoint(checkpoint_path, model, optimizer):
assert os.path.isfile(checkpoint_path)
checkpoint_dict = torch.load(checkpoint_path, map_location='cpu')
iteration = checkpoint_dict['iteration']
optimizer.load_state_dict(checkpoint_dict['optimizer'])
model_for_loading = checkpoint_dict['model']
model.load_state_dict(model_for_loading.state_dict())
print("Loaded checkpoint '{}' (iteration {})" .format(
checkpoint_path, iteration))
return model, optimizer, iteration
def save_checkpoint(model, device, optimizer, learning_rate, iteration, filepath):
print("Saving model and optimizer state at iteration {} to {}".format(
iteration, filepath))
model_for_saving = Wavenet(**wavenet_config).to(device)
model_for_saving.load_state_dict(model.state_dict())
torch.save({'model': model_for_saving,
'iteration': iteration,
'optimizer': optimizer.state_dict(),
'learning_rate': learning_rate}, filepath)
def save_checkpoint_autoencoder(model, device, use_VAE, optimizer, learning_rate, iteration, filepath):
print("Saving model and optimizer state at iteration {} to {}".format(
iteration, filepath))
model_for_saving = WavenetAutoencoder(wavenet_config, cond_wavenet_config, use_VAE).to(device)
model_for_saving.load_state_dict(model.state_dict())
torch.save({'model': model_for_saving,
'iteration': iteration,
'optimizer': optimizer.state_dict(),
'learning_rate': learning_rate}, filepath)
def train(num_gpus, rank, group_name, device, output_directory, epochs, learning_rate,
iters_per_checkpoint, batch_size, seed, checkpoint_path,
use_scheduled_sampling=False,
use_wavenet_autoencoder=False, use_variational_autoencoder=False, diversity_scale=0.005,
use_logistic_mixtures=False, n_mixtures=3,
audio_hz=16000, midi_hz=250):
if num_gpus > 1:
device = init_distributed(rank, num_gpus, group_name, **dist_config)
device = torch.device(device)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
if use_logistic_mixtures:
sampler = DML.SampleDiscretizedMixLogistics()
criterion = DML.DiscretizedMixLogisticLoss()
else:
sampler = utils.CategoricalSampler()
criterion = CrossEntropyLoss()
if use_wavenet_autoencoder:
model = WavenetAutoencoder(wavenet_config, cond_wavenet_config, use_variational_autoencoder).to(device)
if use_variational_autoencoder:
diversity_loss = L2DiversityLoss()
else:
model = Wavenet(**wavenet_config).to(device)
if num_gpus > 1:
model = apply_gradient_allreduce(model)
if use_scheduled_sampling:
scheduled_sampler = ScheduledSamplerWithPatience(model, sampler, **scheduled_sampler_config)
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
# Load checkpoint if one exists
iteration = 0
if checkpoint_path != "":
model, optimizer, iteration = load_checkpoint(checkpoint_path, model, optimizer)
iteration += 1
# Dataloader
trainset = MaestroDataloader(**data_config)
if num_gpus > 1:
train_sampler = DistributedSampler(trainset)
else:
train_sampler = None
train_loader = DataLoader(trainset, num_workers=1, shuffle=False,
sampler=train_sampler,
batch_size=batch_size,
pin_memory=False,
drop_last=True)
# Get shared output_directory ready for distributed
if rank == 0:
if not os.path.isdir(output_directory):
os.makedirs(output_directory)
os.chmod(output_directory, 0o775)
print("output directory", output_directory)
# Initialize training variables
epoch_offset = max(0, int(iteration / len(train_loader)))
start_iter = iteration
loss_idx = 0
loss_sum = 0
print(output_directory)
# write loss to csv file
# FLAG change these to write to the output directory.
loss_writer = DictWriter(open(output_directory + "/train.csv", 'w', newline=''),
fieldnames=['iteration', 'loss'])
loss_writer.writeheader()
signal_writer = DictWriter(open(output_directory + "/signal.csv", "w", newline=''),
fieldnames=['iteration', 'cosim', 'p-dist', 'forwardMagnitude', 'midiMagnitude'])
signal_writer.writeheader()
model.train()
# ================ MAIN TRAINING LOOP! ===================
for epoch in range(epoch_offset, epochs):
print("Epoch: {}".format(epoch))
for i, batch in enumerate(train_loader):
model.zero_grad()
x, y = batch
x = as_variable(x, device)
y = as_variable(y, device)
y_true = y.clone()
if use_scheduled_sampling:
y = scheduled_sampler(x, y)
y_preds = model((x, y))
if use_wavenet_autoencoder:
q_bar = y_preds[1]
y_preds = y_preds[0]
loss = criterion(y_preds, y_true)
if use_variational_autoencoder:
div_loss = diversity_loss(q_bar)
loss = loss + (diversity_scale * div_loss)
if num_gpus > 1:
reduced_loss = reduce_tensor(loss.data, num_gpus).item()
else:
reduced_loss = loss.data.item()
loss.backward()
optimizer.step()
print("total loss: {}:\t{:.9f}".format(iteration, reduced_loss))
if use_variational_autoencoder:
print(" diversity loss: {:.9f}".format(div_loss))
if use_scheduled_sampling:
scheduled_sampler.update(reduced_loss)
# record running average of loss
loss_sum += reduced_loss
loss_idx += 1
if (iteration % 20 == 0):
print("floating avg: " + str(loss_sum/loss_idx))
#loss_writer.writerow({"iteration": str(i),
# "loss": str(reduced_loss)})
loss_sum = 0
loss_idx = 0
# save model
if (iteration % iters_per_checkpoint == 0):
if rank == 0:
checkpoint_path = "{}/wavenet_{}".format(output_directory, iteration)
if use_wavenet_autoencoder:
save_checkpoint_autoencoder(model, device, use_variational_autoencoder, optimizer, learning_rate,
iteration, checkpoint_path)
else:
save_checkpoint(model, device, optimizer, learning_rate, iteration,
checkpoint_path)
iteration += 1
del loss
# end loop
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('-c', '--config', type=str,
help='JSON file for configuration')
parser.add_argument('-r', '--rank', type=int, default=0,
help='rank of process for distributed')
parser.add_argument('-g', '--group_name', type=str, default='',
help='name of group for distributed')
args = parser.parse_args()
# Parse configs. Globals nicer in this case
with open(args.config) as f:
data = f.read()
config = json.loads(data)
train_config = config["train_config"]
global data_config
data_config = config["data_config"]
global scheduled_sampler_config
scheduled_sampler_config = config["scheduled_sampler_config"]
global dist_config
dist_config = config["dist_config"]
global wavenet_config
wavenet_config = config["wavenet_config"]
global cond_wavenet_config
cond_wavenet_config = config["cond_wavenet_config"]
num_gpus = torch.cuda.device_count()
if num_gpus > 1:
if args.group_name == '':
print("WARNING: Multiple GPUs detected but no distributed group set")
print("Only running 1 GPU. Use distributed.py for multiple GPUs")
num_gpus = 1
if num_gpus == 1 and args.rank != 0:
raise Exception("Doing single GPU training on rank > 0")
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = False
train(num_gpus, args.rank, args.group_name, **train_config)