multimodalart HF Staff commited on
Commit
8104203
·
verified ·
1 Parent(s): e1cc9e4

[Admin maintenance] Migrate grant to ZeroGPU

Browse files
Files changed (1) hide show
  1. app.py +37 -36
app.py CHANGED
@@ -1,6 +1,17 @@
 
1
  import torch
2
  torch.manual_seed(160923)
3
 
 
 
 
 
 
 
 
 
 
 
4
 
5
  import gradio as gr
6
  import torch.cuda
@@ -24,14 +35,38 @@ class TTSWebUI:
24
  path_to_iso_list = hf_hub_download(repo_id="Flux9665/ToucanTTS", filename="iso_to_fullname.json")
25
  iso_to_name = load_json_from_path(path_to_iso_list)
26
  text_selection = [f"{iso_to_name[iso_code]} ({iso_code})" for iso_code in iso_to_name]
27
- # accent_selection = [f"{iso_to_name[iso_code]} Accent ({iso_code})" for iso_code in iso_to_name]
28
 
29
  self.controllable_ui = ControllableInterface(gpu_id=gpu_id,
30
  available_artificial_voices=available_artificial_voices,
31
  tts_model_path=tts_model_path,
32
  vocoder_model_path=vocoder_model_path,
33
  embedding_gan_path=embedding_gan_path)
34
- self.iface = gr.Interface(fn=self.read,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
  inputs=[gr.Textbox(lines=2,
36
  placeholder="write what you want the synthesis to read here...",
37
  value="What I cannot create, I do not understand.",
@@ -45,9 +80,6 @@ class TTSWebUI:
45
  gr.Slider(minimum=0, maximum=available_artificial_voices, step=1, value=5, label="Random Seed for the artificial Voice"),
46
  gr.Slider(minimum=-10.0, maximum=10.0, step=0.1, value=0.0, label="Gender of artificial Voice"),
47
  gr.Audio(type="filepath", show_label=True, container=True, label="[OPTIONAL] Voice to Clone (if left empty, will use an artificial voice instead)"),
48
- # gr.Slider(minimum=0.5, maximum=1.5, step=0.1, value=1.0, label="Pitch Variance Scale"),
49
- # gr.Slider(minimum=0.5, maximum=1.5, step=0.1, value=1.0, label="Energy Variance Scale"),
50
- # gr.Slider(minimum=-10.0, maximum=10.0, step=0.1, value=0.0, label="Voice Depth")
51
  ],
52
  outputs=[gr.Audio(type="numpy", label="Speech"),
53
  gr.Image(label="Visualization")],
@@ -57,37 +89,6 @@ class TTSWebUI:
57
  theme=gr.themes.Ocean(primary_hue="amber", secondary_hue="orange"))
58
  self.iface.launch()
59
 
60
- def read(self,
61
- prompt,
62
- language,
63
- prosody_creativity,
64
- duration_scaling_factor,
65
- voice_seed,
66
- emb1,
67
- reference_audio,
68
- # pitch_variance_scale,
69
- # energy_variance_scale,
70
- # emb2
71
- ):
72
- sr, wav, fig = self.controllable_ui.read(prompt,
73
- reference_audio,
74
- language.split(" ")[-1].split("(")[1].split(")")[0],
75
- language.split(" ")[-1].split("(")[1].split(")")[0],
76
- voice_seed,
77
- prosody_creativity,
78
- duration_scaling_factor,
79
- 1.,
80
- 1.0,
81
- 1.0,
82
- emb1,
83
- 0.,
84
- 0.,
85
- 0.,
86
- 0.,
87
- 0.,
88
- -24.)
89
- return (sr, float2pcm(wav)), fig
90
-
91
 
92
  if __name__ == '__main__':
93
  TTSWebUI(gpu_id="cuda" if torch.cuda.is_available() else "cpu")
 
1
+ import spaces
2
  import torch
3
  torch.manual_seed(160923)
4
 
5
+ _orig_set_device = torch.cuda.set_device
6
+ def _safe_set_device(device):
7
+ try:
8
+ _orig_set_device(device)
9
+ except RuntimeError as exc:
10
+ if "Low-level CUDA init" in str(exc):
11
+ return
12
+ raise
13
+ torch.cuda.set_device = _safe_set_device
14
+
15
 
16
  import gradio as gr
17
  import torch.cuda
 
35
  path_to_iso_list = hf_hub_download(repo_id="Flux9665/ToucanTTS", filename="iso_to_fullname.json")
36
  iso_to_name = load_json_from_path(path_to_iso_list)
37
  text_selection = [f"{iso_to_name[iso_code]} ({iso_code})" for iso_code in iso_to_name]
 
38
 
39
  self.controllable_ui = ControllableInterface(gpu_id=gpu_id,
40
  available_artificial_voices=available_artificial_voices,
41
  tts_model_path=tts_model_path,
42
  vocoder_model_path=vocoder_model_path,
43
  embedding_gan_path=embedding_gan_path)
44
+
45
+ controllable_ui = self.controllable_ui
46
+
47
+ @spaces.GPU
48
+ def read(prompt, language, prosody_creativity, duration_scaling_factor, voice_seed, emb1, reference_audio):
49
+ iso = language.split(" ")[-1].split("(")[1].split(")")[0]
50
+ sr, wav, fig = controllable_ui.read(prompt,
51
+ reference_audio,
52
+ iso,
53
+ iso,
54
+ voice_seed,
55
+ prosody_creativity,
56
+ duration_scaling_factor,
57
+ 1.,
58
+ 1.0,
59
+ 1.0,
60
+ emb1,
61
+ 0.,
62
+ 0.,
63
+ 0.,
64
+ 0.,
65
+ 0.,
66
+ -24.)
67
+ return (sr, float2pcm(wav)), fig
68
+
69
+ self.iface = gr.Interface(fn=read,
70
  inputs=[gr.Textbox(lines=2,
71
  placeholder="write what you want the synthesis to read here...",
72
  value="What I cannot create, I do not understand.",
 
80
  gr.Slider(minimum=0, maximum=available_artificial_voices, step=1, value=5, label="Random Seed for the artificial Voice"),
81
  gr.Slider(minimum=-10.0, maximum=10.0, step=0.1, value=0.0, label="Gender of artificial Voice"),
82
  gr.Audio(type="filepath", show_label=True, container=True, label="[OPTIONAL] Voice to Clone (if left empty, will use an artificial voice instead)"),
 
 
 
83
  ],
84
  outputs=[gr.Audio(type="numpy", label="Speech"),
85
  gr.Image(label="Visualization")],
 
89
  theme=gr.themes.Ocean(primary_hue="amber", secondary_hue="orange"))
90
  self.iface.launch()
91
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92
 
93
  if __name__ == '__main__':
94
  TTSWebUI(gpu_id="cuda" if torch.cuda.is_available() else "cpu")