Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torchaudio
|
3 |
+
import gradio as gr
|
4 |
+
|
5 |
+
device="cpu"
|
6 |
+
bundle = torchaudio.pipelines.TACOTRON2_WAVERNN_PHONE_LJSPEECH
|
7 |
+
processor = bundle.get_text_processor()
|
8 |
+
tacotron2 = bundle.get_tacotron2().to(device)
|
9 |
+
|
10 |
+
# Workaround to load model mapped on GPU
|
11 |
+
# https://stackoverflow.com/a/61840832
|
12 |
+
waveglow = torch.hub.load(
|
13 |
+
"NVIDIA/DeepLearningExamples:torchhub",
|
14 |
+
"nvidia_waveglow",
|
15 |
+
model_math="fp32",
|
16 |
+
pretrained=False,
|
17 |
+
)
|
18 |
+
checkpoint = torch.hub.load_state_dict_from_url(
|
19 |
+
"https://api.ngc.nvidia.com/v2/models/nvidia/waveglowpyt_fp32/versions/1/files/nvidia_waveglowpyt_fp32_20190306.pth", # noqa: E501
|
20 |
+
progress=False,
|
21 |
+
map_location=device,
|
22 |
+
)
|
23 |
+
state_dict = {key.replace("module.", ""): value for key, value in checkpoint["state_dict"].items()}
|
24 |
+
|
25 |
+
waveglow.load_state_dict(state_dict)
|
26 |
+
waveglow = waveglow.remove_weightnorm(waveglow)
|
27 |
+
waveglow = waveglow.to(device)
|
28 |
+
waveglow.eval()
|
29 |
+
|
30 |
+
def inference(text):
|
31 |
+
|
32 |
+
with torch.inference_mode():
|
33 |
+
processed, lengths = processor(text)
|
34 |
+
processed = processed.to(device)
|
35 |
+
lengths = lengths.to(device)
|
36 |
+
spec, _, _ = tacotron2.infer(processed, lengths)
|
37 |
+
|
38 |
+
|
39 |
+
|
40 |
+
with torch.no_grad():
|
41 |
+
waveforms = waveglow.infer(spec)
|
42 |
+
|
43 |
+
torchaudio.save("output_waveglow.wav", waveforms[0:1].cpu(), sample_rate=22050)
|
44 |
+
return "output_waveglow.wav"
|
45 |
+
|
46 |
+
title="TACOTRON 2"
|
47 |
+
description="Gradio demo for TACOTRON 2: The Tacotron 2 model for generating mel spectrograms from text. To use it, simply add you text or click on one of the examples to load them. Read more at the links below."
|
48 |
+
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/1712.05884' target='_blank'>Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions</a> | <a href='https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/SpeechSynthesis/Tacotron2' target='_blank'>Github Repo</a></p>"
|
49 |
+
examples=[["life is like a box of chocolates"]]
|
50 |
+
gr.Interface(inference,"text",gr.outputs.Audio(type="file"),title=title,description=description,article=article,examples=examples).launch(enable_queue=True)
|