AmbientAudioGAN/train_declipper.py at master · shehzeen/AmbientAudioGAN · GitHub

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
import time
import pickle

import numpy as np
import tensorflow as tf

from AudioModel.loader_measured import waveform_decoder
from AudioModel.model import Modes
from AudioModel.util import override_model_attrs
from DeclippingModel import WaveAE

def train(fps, args):
  # Initialize model
  model = WaveAE(Modes.TRAIN)
  model, summary = override_model_attrs(model, args.model_overrides)
  print('-' * 80)
  print(summary)
  print('-' * 80)

  # Load data
  with tf.name_scope('loader'):
    clean, x = waveform_decoder(
        fps=fps,
        batch_size=model.train_batch_size,
        subseq_len=model.subseq_len,
        audio_fs=model.audio_fs,
        audio_mono=True,
        audio_normalize=True,
        decode_fastwav=args.data_fastwav,
        decode_parallel_calls=4,
        repeat=True,
        shuffle=True,
        shuffle_buffer_size=4096,
        subseq_randomize_offset=args.data_randomize_offset,
        subseq_overlap_ratio=args.data_overlap_ratio,
        subseq_pad_end=True,
        prefetch_size=64,
        gpu_num=0)

  # Create model
  model(clean, x)

  # Train
  # model_dir_path = "/data2/paarth/TrainDir/WaveAE/WaveAEsc09_l1batchnormFalse/eval_sc09_valid"
  # ckpt = 253802
  with tf.train.MonitoredTrainingSession(
      checkpoint_dir=args.train_dir,
      save_checkpoint_secs=args.train_ckpt_every_nsecs,
      save_summaries_secs=args.train_summary_every_nsecs) as sess:
    while not sess.should_stop():
      model.train_loop(sess)

def eval(fps, args):
  eval_dir = os.path.join(args.train_dir, 'eval_valid')
  if not os.path.isdir(eval_dir):
    os.makedirs(eval_dir)

  model = WaveAE(Modes.EVAL)
  model, summary = override_model_attrs(model, args.model_overrides)
  print('-' * 80)
  print(summary)
  print('-' * 80)

  # Load data
  with tf.name_scope('loader'):
    clean, x = waveform_decoder(
        fps=fps,
        batch_size=model.eval_batch_size,
        subseq_len=model.subseq_len,
        audio_fs=model.audio_fs,
        audio_mono=True,
        audio_normalize=True,
        decode_fastwav=args.data_fastwav,
        decode_parallel_calls=1,
        repeat=False,
        shuffle=False,
        shuffle_buffer_size=None,
        subseq_randomize_offset=False,
        subseq_overlap_ratio=0.,
        subseq_pad_end=True,
        prefetch_size=None,
        gpu_num=None)

  model.build_denoiser(clean, x)

  saver = tf.train.Saver(var_list=model.restore_vars, max_to_keep=1)
  summary_writer = tf.summary.FileWriter(eval_dir)

  ckpt_fp = None
  while True:
    latest_ckpt_fp = tf.train.latest_checkpoint(args.train_dir)
    if latest_ckpt_fp != ckpt_fp:
      ckpt_fp = latest_ckpt_fp
      print('Evaluating {}'.format(ckpt_fp))
      with tf.Session() as sess:
        model.eval_ckpt(ckpt_fp, sess, summary_writer, saver, eval_dir)
      print('Done!')
    time.sleep(1)

if __name__ == '__main__':
  from argparse import ArgumentParser
  import glob
  import os

  parser = ArgumentParser()

  parser.add_argument('mode', type=str, choices=['train', 'eval'])
  parser.add_argument('train_dir', type=str)

  parser.add_argument('--data_dir', type=str, required=True)
  parser.add_argument('--data_fastwav', dest='data_fastwav', action='store_true')
  parser.add_argument('--data_randomize_offset', dest='data_randomize_offset', action='store_true')
  parser.add_argument('--data_overlap_ratio', type=float)

  parser.add_argument('--model_overrides', type=str)

  parser.add_argument('--train_ckpt_every_nsecs', type=int)
  parser.add_argument('--train_summary_every_nsecs', type=int)


  parser.set_defaults(
      mode=None,
      train_dir=None,
      data_dir=None,
      data_fastwav=False,
      data_randomize_offset=False,
      data_overlap_ratio=0.,
      model_cfg_overrides=None,
      train_ckpt_every_nsecs=360,
      train_summary_every_nsecs=60,
      )

  args = parser.parse_args()

  if not os.path.isdir(args.train_dir):
    os.makedirs(args.train_dir)

  fps = glob.glob(os.path.join(args.data_dir, '*_clean.wav'))
  print('Found {} audio files'.format(len(fps)))

  if args.mode == 'train':
    train(fps, args)
  elif args.mode == 'eval':
    eval(fps, args)
  else:
    raise NotImplementedError()