Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

print average text length, fix for Nancy preprocessor

This commit is contained in:
Eren Golge 2018-11-20 12:54:33 +01:00
parent b8ca19fd2c
commit 660f8c7b66
2 changed files with 11 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
datasets/preprocess.py
View File

@ -55,6 +55,6 @@ def nancy(root_path, meta_file):
id = line.split()[1] id = line.split()[1]
text = line[line.find('"')+1:line.rfind('"')-1] text = line[line.find('"')+1:line.rfind('"')-1]
wav_file = root_path + 'wavn/' + id + '.wav' wav_file = root_path + 'wavn/' + id + '.wav'
items.append(text, wav_file) items.append([text, wav_file])
random.shuffle(items) random.shuffle(items)
return items return items

17
train.py
View File

@ -51,6 +51,7 @@ def train(model, criterion, criterion_st, data_loader, optimizer, optimizer_st,
mel_input = data[3] mel_input = data[3]
mel_lengths = data[4] mel_lengths = data[4]
stop_targets = data[5] stop_targets = data[5]
avg_text_length = torch.mean(text_lengths.float())
# set stop targets view, we predict a single stop token per r frames prediction # set stop targets view, we predict a single stop token per r frames prediction
stop_targets = stop_targets.view(text_input.shape[0], stop_targets = stop_targets.view(text_input.shape[0],
@ -68,12 +69,12 @@ def train(model, criterion, criterion_st, data_loader, optimizer, optimizer_st,
# dispatch data to GPU # dispatch data to GPU
if use_cuda: if use_cuda:
text_input = text_input.cuda() text_input = text_input.cuda(non_blocking=True)
text_lengths = text_lengths.cuda() text_lengths = text_lengths.cuda(non_blocking=True)
mel_input = mel_input.cuda() mel_input = mel_input.cuda(non_blocking=True)
mel_lengths = mel_lengths.cuda() mel_lengths = mel_lengths.cuda(non_blocking=True)
linear_input = linear_input.cuda() linear_input = linear_input.cuda(non_blocking=True)
stop_targets = stop_targets.cuda() stop_targets = stop_targets.cuda(non_blocking=True)
# compute mask for padding # compute mask for padding
mask = sequence_mask(text_lengths) mask = sequence_mask(text_lengths)
@ -129,10 +130,10 @@ def train(model, criterion, criterion_st, data_loader, optimizer, optimizer_st,
print( print(
" | > Step:{}/{} GlobalStep:{} TotalLoss:{:.5f} LinearLoss:{:.5f} " " | > Step:{}/{} GlobalStep:{} TotalLoss:{:.5f} LinearLoss:{:.5f} "
"MelLoss:{:.5f} StopLoss:{:.5f} GradNorm:{:.5f} " "MelLoss:{:.5f} StopLoss:{:.5f} GradNorm:{:.5f} "
"GradNormST:{:.5f} StepTime:{:.2f} LR:{:.6f}".format( "GradNormST:{:.5f} AvgTextLen:{:.1f} StepTime:{:.2f} LR:{:.6f}".format(
num_iter, batch_n_iter, current_step, loss.item(), num_iter, batch_n_iter, current_step, loss.item(),
linear_loss.item(), mel_loss.item(), stop_loss.item(), linear_loss.item(), mel_loss.item(), stop_loss.item(),
grad_norm, grad_norm_st, step_time, current_lr), grad_norm, grad_norm_st, avg_text_length, step_time, current_lr),
flush=True) flush=True)
avg_linear_loss += linear_loss.item() avg_linear_loss += linear_loss.item()