Spaces:
Running
on
Zero
Running
on
Zero
Upload app.py
Browse files
app.py
CHANGED
@@ -13,6 +13,30 @@ import spaces
|
|
13 |
import torch
|
14 |
import yaml
|
15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
random_texts = {}
|
17 |
for lang in ['en', 'ja']:
|
18 |
with open(f'{lang}.txt', 'r') as r:
|
@@ -86,25 +110,6 @@ VOCAB = get_vocab()
|
|
86 |
def tokenize(ps):
|
87 |
return [i for i in map(VOCAB.get, ps) if i is not None]
|
88 |
|
89 |
-
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
90 |
-
|
91 |
-
snapshot = snapshot_download(repo_id='hexgrad/kokoro', allow_patterns=['*.pt', '*.pth', '*.yml'], use_auth_token=os.environ['TOKEN'])
|
92 |
-
config = yaml.safe_load(open(os.path.join(snapshot, 'config.yml')))
|
93 |
-
model = build_model(config['model_params'])
|
94 |
-
for key, value in model.items():
|
95 |
-
for module in value.children():
|
96 |
-
if isinstance(module, torch.nn.RNNBase):
|
97 |
-
module.flatten_parameters()
|
98 |
-
_ = [model[key].eval() for key in model]
|
99 |
-
_ = [model[key].to(device) for key in model]
|
100 |
-
for key, state_dict in torch.load(os.path.join(snapshot, 'net.pth'), map_location='cpu', weights_only=True)['net'].items():
|
101 |
-
assert key in model, key
|
102 |
-
try:
|
103 |
-
model[key].load_state_dict(state_dict)
|
104 |
-
except:
|
105 |
-
state_dict = {k[7:]: v for k, v in state_dict.items()}
|
106 |
-
model[key].load_state_dict(state_dict, strict=False)
|
107 |
-
|
108 |
CHOICES = {
|
109 |
'🇺🇸 🚺 American Female 0': 'af_0',
|
110 |
'🇺🇸 🚺 Bella': 'af_bella',
|
|
|
13 |
import torch
|
14 |
import yaml
|
15 |
|
16 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
17 |
+
|
18 |
+
snapshot = snapshot_download(repo_id='hexgrad/kokoro', allow_patterns=['*.pt', '*.pth', '*.yml'], use_auth_token=os.environ['TOKEN'])
|
19 |
+
config = yaml.safe_load(open(os.path.join(snapshot, 'config.yml')))
|
20 |
+
model = build_model(config['model_params'])
|
21 |
+
for key, value in model.items():
|
22 |
+
for module in value.children():
|
23 |
+
if isinstance(module, torch.nn.RNNBase):
|
24 |
+
module.flatten_parameters()
|
25 |
+
|
26 |
+
_ = [model[key].eval() for key in model]
|
27 |
+
_ = [model[key].to(device) for key in model]
|
28 |
+
for key, state_dict in torch.load(os.path.join(snapshot, 'net.pth'), map_location='cpu', weights_only=True)['net'].items():
|
29 |
+
assert key in model, key
|
30 |
+
try:
|
31 |
+
model[key].load_state_dict(state_dict)
|
32 |
+
except:
|
33 |
+
state_dict = {k[7:]: v for k, v in state_dict.items()}
|
34 |
+
model[key].load_state_dict(state_dict, strict=False)
|
35 |
+
|
36 |
+
PARAM_COUNT = sum(p.numel() for value in model.values() for p in value.parameters())
|
37 |
+
print('PARAM_COUNT', PARAM_COUNT)
|
38 |
+
assert PARAM_COUNT < 82_000_000, PARAM_COUNT
|
39 |
+
|
40 |
random_texts = {}
|
41 |
for lang in ['en', 'ja']:
|
42 |
with open(f'{lang}.txt', 'r') as r:
|
|
|
110 |
def tokenize(ps):
|
111 |
return [i for i in map(VOCAB.get, ps) if i is not None]
|
112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
113 |
CHOICES = {
|
114 |
'🇺🇸 🚺 American Female 0': 'af_0',
|
115 |
'🇺🇸 🚺 Bella': 'af_bella',
|