Spaces:
Running
Running
Add model info
Browse files- app.py +40 -12
- radtts-pp-dap-model.json → config.json +0 -0
app.py
CHANGED
@@ -3,9 +3,10 @@ import sys
|
|
3 |
import json
|
4 |
import time
|
5 |
|
6 |
-
from
|
7 |
-
from
|
8 |
from enum import Enum
|
|
|
9 |
|
10 |
try:
|
11 |
import spaces # it's for ZeroGPU
|
@@ -26,7 +27,7 @@ from torch_env import device
|
|
26 |
import gradio as gr
|
27 |
|
28 |
|
29 |
-
#
|
30 |
from vocos import Vocos
|
31 |
|
32 |
use_zerogpu = False
|
@@ -70,13 +71,10 @@ download_file_from_repo(
|
|
70 |
)
|
71 |
|
72 |
# Init the model
|
73 |
-
config = "radtts-pp-dap-model.json"
|
74 |
-
radtts_path = "models/radtts-pp-dap-model/model_dap_84000_state.pt"
|
75 |
-
|
76 |
params = []
|
77 |
|
78 |
# Load the config
|
79 |
-
with open(config) as f:
|
80 |
data = f.read()
|
81 |
|
82 |
config = json.loads(data)
|
@@ -86,18 +84,33 @@ data_config = config["data_config"]
|
|
86 |
model_config = config["model_config"]
|
87 |
|
88 |
# Load vocoder
|
89 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
|
91 |
# Load RAD-TTS
|
92 |
radtts = RADTTS(**model_config).to(device)
|
93 |
-
|
94 |
radtts.enable_inverse_cache() # cache inverse matrix for 1x1 invertible convs
|
95 |
|
96 |
-
|
97 |
-
|
|
|
|
|
|
|
|
|
98 |
radtts.eval()
|
99 |
|
100 |
-
|
|
|
|
|
|
|
|
|
101 |
|
102 |
ignore_keys = ["training_files", "validation_files"]
|
103 |
text_processor = TextProcessor(
|
@@ -142,6 +155,21 @@ tech_env = f"""
|
|
142 |
|
143 |
- Python: {sys.version}
|
144 |
- Torch device: {device}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
145 |
""".strip()
|
146 |
|
147 |
tech_libraries = f"""
|
|
|
3 |
import json
|
4 |
import time
|
5 |
|
6 |
+
from os.path import getsize
|
7 |
+
from pathlib import Path
|
8 |
from enum import Enum
|
9 |
+
from importlib.metadata import version, PackageNotFoundError
|
10 |
|
11 |
try:
|
12 |
import spaces # it's for ZeroGPU
|
|
|
27 |
import gradio as gr
|
28 |
|
29 |
|
30 |
+
# Vocoder
|
31 |
from vocos import Vocos
|
32 |
|
33 |
use_zerogpu = False
|
|
|
71 |
)
|
72 |
|
73 |
# Init the model
|
|
|
|
|
|
|
74 |
params = []
|
75 |
|
76 |
# Load the config
|
77 |
+
with open("config.json") as f:
|
78 |
data = f.read()
|
79 |
|
80 |
config = json.loads(data)
|
|
|
84 |
model_config = config["model_config"]
|
85 |
|
86 |
# Load vocoder
|
87 |
+
vocos_config = hf_hub_download("patriotyk/vocos-mel-hifigan-compat-44100khz", "config.yaml")
|
88 |
+
vocos_model = hf_hub_download("patriotyk/vocos-mel-hifigan-compat-44100khz", "pytorch_model.bin")
|
89 |
+
|
90 |
+
vocos_model_path = Path(vocos_model)
|
91 |
+
state_dict = torch.load(vocos_model_path, map_location="cpu")
|
92 |
+
|
93 |
+
vocos = Vocos.from_hparams(vocos_config).to(device)
|
94 |
+
vocos.load_state_dict(state_dict, strict=True)
|
95 |
+
vocos.eval()
|
96 |
|
97 |
# Load RAD-TTS
|
98 |
radtts = RADTTS(**model_config).to(device)
|
|
|
99 |
radtts.enable_inverse_cache() # cache inverse matrix for 1x1 invertible convs
|
100 |
|
101 |
+
radtts_model_path = Path("models/radtts-pp-dap-model/model_dap_84000_state.pt")
|
102 |
+
|
103 |
+
checkpoint_dict = torch.load(radtts_model_path, map_location="cpu")
|
104 |
+
state_dict = checkpoint_dict["state_dict"]
|
105 |
+
|
106 |
+
radtts.load_state_dict(state_dict, strict=False)
|
107 |
radtts.eval()
|
108 |
|
109 |
+
radtts_params = f'{sum(param.numel() for param in radtts.parameters()):,}'
|
110 |
+
vocos_params = f'{sum(param.numel() for param in vocos.parameters()):,}'
|
111 |
+
|
112 |
+
print(f"Loaded checkpoint (RAD-TTS++), number of parameters: {radtts_params}")
|
113 |
+
print(f"Loaded checkpoint (Vocos), number of parameters: {vocos_params}")
|
114 |
|
115 |
ignore_keys = ["training_files", "validation_files"]
|
116 |
text_processor = TextProcessor(
|
|
|
155 |
|
156 |
- Python: {sys.version}
|
157 |
- Torch device: {device}
|
158 |
+
|
159 |
+
#### Model
|
160 |
+
|
161 |
+
##### Acoustic model
|
162 |
+
|
163 |
+
- Name: RAD-TTS++ (DAP)
|
164 |
+
- Parameters: {radtts_params}
|
165 |
+
- File size: {getsize(radtts_model_path) / 1e6:.2f} MB
|
166 |
+
|
167 |
+
##### Vocoder
|
168 |
+
|
169 |
+
- Name: Vocos
|
170 |
+
- Parameters: {vocos_params}
|
171 |
+
- File size: {getsize(vocos_model_path) / 1e6:.2f} MB
|
172 |
+
|
173 |
""".strip()
|
174 |
|
175 |
tech_libraries = f"""
|
radtts-pp-dap-model.json → config.json
RENAMED
File without changes
|