sess.run([loss],
feed_dict={
train_flag: True,
x: batch[1],
y_true: batch[0]
}))
print("Iteration {}, Loss {}".format(i + 1, loss_val))
val_loss_file.close()
train_loss_file.close()
# Save the variables to disk.
save_path = saver.save(
sess, "aux/model_checkpoints/{}_final.ckpt".format(model_name))
print("Model checkpoints will be saved in file: {}".format(save_path))
truth, example = read_file_pair(val_truth_ds_pairs[1])
y_reco = model.eval(feed_dict={
train_flag: False,
x: example.reshape(1, -1, 1)
},
session=sess).flatten()
print('difference between truth and example (first 20 elements)')
print(truth.flatten()[:20] - example.flatten()[:20])
print('difference between truth and reconstruction (first 20 elements)')
print(truth.flatten()[:20] - y_reco[:20])
print('writting output audio files')
librosa.output.write_wav('full_train_validation_true.wav',
y=truth.flatten(),
sr=true_br)
print("Epoch {}, Loss {}".format((i + 1), loss_val))
train_loss_file.write('{}\n'.format(loss_val))
if write_tb:
summary = sess.run([merged],
feed_dict={
train_flag: True,
x: batch[1],
y_true: batch[0]
})
train_writer.add_summary(summary, i)
save_path = saver.save(sess, "aux/model_checkpoints/overtrain_final.ckpt")
print("Model checkpoints will be saved in file: {}".format(save_path))
train_loss_file.close()
truth, example = read_file_pair(train_truth_ds_pairs[example_number])
y_reco = model.eval(feed_dict={
train_flag: True,
x: example.reshape(1, -1, 1)
},
session=sess).flatten()
print('difference between truth and example (first 20 elements)')
print(truth.flatten()[:20] - example.flatten()[:20])
print('difference between truth and reconstruction (first 20 elements)')
print(truth.flatten()[:20] - y_reco[:20])
print('writting output audio files')
librosa.output.write_wav('overtrain_true.wav', y=truth.flatten(), sr=true_br)
librosa.output.write_wav('overtrain_ds.wav', y=example.flatten(), sr=true_br)
test_loss_file = open('test_loss.txt', 'w')
count = 0
for pair in next_batch(BATCH_SIZE, test_truth_ds_pairs):
loss_test = sess.run([waveform_mse],
feed_dict={train_flag: False,
x: pair[1],
y_true: pair[0]}
)
test_loss_file.write('{}\n'.format(np.mean(loss_test)))
print("Iteration {}, Test Loss {}".format((count + 1), loss_test))
count += 1
test_loss_file.close()
truth, example = read_file_pair(test_truth_ds_pairs[0])
y_reco = model.eval(feed_dict={train_flag: False,
x: example.reshape(1, -1, 1)},
session=sess).flatten()
print('difference between truth and example (first 20 elements)')
print(truth.flatten()[:20] - example.flatten()[:20])
print('difference between truth and reconstruction (first 20 elements)')
print(truth.flatten()[:20] - y_reco[:20])
# if waveform_reduction_factor == 1:
print('writting output audio files')
librosa.output.write_wav('full_train_test_true.wav',
y=truth.flatten(), sr=true_br)
librosa.output.write_wav('full_train_test_ds.wav',
y=example.flatten(), sr=true_br)