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

Plot specs and alignments for test sentences

This commit is contained in:
Eren G 2018-07-23 16:14:46 +02:00
parent 766e88700d
commit c72f309068
5 changed files with 19 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
.compute
View File

@ -1,3 +1,6 @@
#!/bin/bash #!/bin/bash
source ../tmp/venv/bin/activate # source ../tmp/venv/bin/activate
# ls /snakepit/jobs/650/keep/
source /snakepit/jobs/650/keep/venv/bin/activate
# source /snakepit/jobs/560/tmp/venv/bin/activate
python train.py --config_path config.json --debug true python train.py --config_path config.json --debug true

6
.install
View File

@ -1,4 +1,4 @@
#!/bin/bash #!/bin/bash
virtualenv -p python3 ../tmp/venv # virtualenv -p python3 ../tmp/venv
source ../tmp/venv/bin/activate # source ../tmp/venv/bin/activate
python setup.py develop # python setup.py develop

4
config.json
View File

@ -23,10 +23,10 @@
"griffin_lim_iters": 60, "griffin_lim_iters": 60,
"power": 1.5, "power": 1.5,
"num_loader_workers": 4, "num_loader_workers": 14,
"checkpoint": true, "checkpoint": true,
"save_step": 376, "save_step": 750,
"print_step": 10, "print_step": 10,
"run_eval": false, "run_eval": false,
"data_path": "/snakepit/shared/data/keithito/LJSpeech-1.1/", "data_path": "/snakepit/shared/data/keithito/LJSpeech-1.1/",

10
train.py
View File

@ -302,14 +302,20 @@ def evaluate(model, criterion, criterion_st, data_loader, ap, current_step):
# test sentences # test sentences
ap.griffin_lim_iters = 60 ap.griffin_lim_iters = 60
for idx, test_sentence in enumerate(test_sentences): for idx, test_sentence in enumerate(test_sentences):
wav = synthesis(model, ap, test_sentence, use_cuda, wav, linear_out, alignments = synthesis(model, ap, test_sentence, use_cuda,
c.text_cleaner) c.text_cleaner)
try: try:
wav_name = 'TestSentences/{}'.format(idx) wav_name = 'TestSentences/{}'.format(idx)
tb.add_audio(wav_name, wav, current_step, tb.add_audio(wav_name, wav, current_step,
sample_rate=c.sample_rate) sample_rate=c.sample_rate)
except: except:
pass pass
align_img = alignments[0].data.cpu().numpy()
linear_spec = linear_output[0].data.cpu().numpy()
linear_spec = plot_spectrogram(linear_spec, ap)
align_img = plot_alignment(align_img)
tb.add_image('TestSentences/{}_GroundTruth'.format(idx), gt_spec, current_step)
tb.add_image('TestSentences/{}_Alignment'.format(idx), align_img, current_step)
return avg_linear_loss return avg_linear_loss

4
utils/generic_utils.py
View File

@ -166,7 +166,7 @@ def synthesis(model, ap, text, use_cuda, text_cleaner):
chars_var = torch.from_numpy(seq).unsqueeze(0) chars_var = torch.from_numpy(seq).unsqueeze(0)
if use_cuda: if use_cuda:
chars_var = chars_var.cuda().long() chars_var = chars_var.cuda().long()
_, linear_out, _, _ = model.forward(chars_var) _, linear_out, alignments, _ = model.forward(chars_var)
linear_out = linear_out[0].data.cpu().numpy() linear_out = linear_out[0].data.cpu().numpy()
wav = ap.inv_spectrogram(linear_out.T) wav = ap.inv_spectrogram(linear_out.T)
return wav return wav, linear_out, alignments