Spaces:
Runtime error
Runtime error
xj
commited on
Commit
·
4c9e450
1
Parent(s):
8f95475
[bug] 加快编译速度
Browse files- app.py +2 -4
- requirements.txt +6 -6
- utils.py +53 -53
app.py
CHANGED
|
@@ -22,8 +22,6 @@ from models import SynthesizerTrn
|
|
| 22 |
from text import text_to_sequence
|
| 23 |
import torch
|
| 24 |
from torch import no_grad, LongTensor
|
| 25 |
-
import webbrowser
|
| 26 |
-
import gradio.processing_utils as gr_processing_utils
|
| 27 |
from gradio_client import utils as client_utils
|
| 28 |
limitation = os.getenv("SYSTEM") == "spaces" # limit text and audio length in huggingface spaces
|
| 29 |
|
|
@@ -45,10 +43,10 @@ def get_text(text, hps):
|
|
| 45 |
def vits(text, language, speaker_id, noise_scale, noise_scale_w, length_scale):
|
| 46 |
start = time.perf_counter()
|
| 47 |
if not len(text):
|
| 48 |
-
return
|
| 49 |
text = text.replace('\n', ' ').replace('\r', '').replace(" ", "")
|
| 50 |
if len(text) > 200 and limitation:
|
| 51 |
-
return
|
| 52 |
if language == "中文":
|
| 53 |
text = f"[ZH]{text}[ZH]"
|
| 54 |
elif language == "日文":
|
|
|
|
| 22 |
from text import text_to_sequence
|
| 23 |
import torch
|
| 24 |
from torch import no_grad, LongTensor
|
|
|
|
|
|
|
| 25 |
from gradio_client import utils as client_utils
|
| 26 |
limitation = os.getenv("SYSTEM") == "spaces" # limit text and audio length in huggingface spaces
|
| 27 |
|
|
|
|
| 43 |
def vits(text, language, speaker_id, noise_scale, noise_scale_w, length_scale):
|
| 44 |
start = time.perf_counter()
|
| 45 |
if not len(text):
|
| 46 |
+
return None
|
| 47 |
text = text.replace('\n', ' ').replace('\r', '').replace(" ", "")
|
| 48 |
if len(text) > 200 and limitation:
|
| 49 |
+
return None
|
| 50 |
if language == "中文":
|
| 51 |
text = f"[ZH]{text}[ZH]"
|
| 52 |
elif language == "日文":
|
requirements.txt
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
-
Cython
|
| 2 |
librosa
|
| 3 |
-
matplotlib
|
| 4 |
numpy
|
| 5 |
-
phonemizer
|
| 6 |
-
scipy
|
| 7 |
-
tensorboard
|
| 8 |
torch
|
| 9 |
-
torchvision
|
| 10 |
Unidecode
|
| 11 |
pyopenjtalk
|
| 12 |
ffmpeg
|
|
|
|
| 1 |
+
#Cython
|
| 2 |
librosa
|
| 3 |
+
#matplotlib
|
| 4 |
numpy
|
| 5 |
+
#phonemizer
|
| 6 |
+
#scipy
|
| 7 |
+
#tensorboard
|
| 8 |
torch
|
| 9 |
+
#torchvision
|
| 10 |
Unidecode
|
| 11 |
pyopenjtalk
|
| 12 |
ffmpeg
|
utils.py
CHANGED
|
@@ -42,59 +42,59 @@ def load_checkpoint(checkpoint_path, model, optimizer=None):
|
|
| 42 |
return model, optimizer, learning_rate, iteration
|
| 43 |
|
| 44 |
|
| 45 |
-
def plot_spectrogram_to_numpy(spectrogram):
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
def plot_alignment_to_numpy(alignment, info=None):
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
|
| 99 |
|
| 100 |
def load_audio_to_torch(full_path, target_sampling_rate):
|
|
|
|
| 42 |
return model, optimizer, learning_rate, iteration
|
| 43 |
|
| 44 |
|
| 45 |
+
# def plot_spectrogram_to_numpy(spectrogram):
|
| 46 |
+
# global MATPLOTLIB_FLAG
|
| 47 |
+
# if not MATPLOTLIB_FLAG:
|
| 48 |
+
# import matplotlib
|
| 49 |
+
# matplotlib.use("Agg")
|
| 50 |
+
# MATPLOTLIB_FLAG = True
|
| 51 |
+
# mpl_logger = logging.getLogger('matplotlib')
|
| 52 |
+
# mpl_logger.setLevel(logging.WARNING)
|
| 53 |
+
# import matplotlib.pylab as plt
|
| 54 |
+
# import numpy as np
|
| 55 |
+
|
| 56 |
+
# fig, ax = plt.subplots(figsize=(10,2))
|
| 57 |
+
# im = ax.imshow(spectrogram, aspect="auto", origin="lower",
|
| 58 |
+
# interpolation='none')
|
| 59 |
+
# plt.colorbar(im, ax=ax)
|
| 60 |
+
# plt.xlabel("Frames")
|
| 61 |
+
# plt.ylabel("Channels")
|
| 62 |
+
# plt.tight_layout()
|
| 63 |
+
|
| 64 |
+
# fig.canvas.draw()
|
| 65 |
+
# data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
|
| 66 |
+
# data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
|
| 67 |
+
# plt.close()
|
| 68 |
+
# return data
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
# def plot_alignment_to_numpy(alignment, info=None):
|
| 72 |
+
# global MATPLOTLIB_FLAG
|
| 73 |
+
# if not MATPLOTLIB_FLAG:
|
| 74 |
+
# import matplotlib
|
| 75 |
+
# matplotlib.use("Agg")
|
| 76 |
+
# MATPLOTLIB_FLAG = True
|
| 77 |
+
# mpl_logger = logging.getLogger('matplotlib')
|
| 78 |
+
# mpl_logger.setLevel(logging.WARNING)
|
| 79 |
+
# import matplotlib.pylab as plt
|
| 80 |
+
# import numpy as np
|
| 81 |
+
|
| 82 |
+
# fig, ax = plt.subplots(figsize=(6, 4))
|
| 83 |
+
# im = ax.imshow(alignment.transpose(), aspect='auto', origin='lower',
|
| 84 |
+
# interpolation='none')
|
| 85 |
+
# fig.colorbar(im, ax=ax)
|
| 86 |
+
# xlabel = 'Decoder timestep'
|
| 87 |
+
# if info is not None:
|
| 88 |
+
# xlabel += '\n\n' + info
|
| 89 |
+
# plt.xlabel(xlabel)
|
| 90 |
+
# plt.ylabel('Encoder timestep')
|
| 91 |
+
# plt.tight_layout()
|
| 92 |
+
|
| 93 |
+
# fig.canvas.draw()
|
| 94 |
+
# data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
|
| 95 |
+
# data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
|
| 96 |
+
# plt.close()
|
| 97 |
+
# return data
|
| 98 |
|
| 99 |
|
| 100 |
def load_audio_to_torch(full_path, target_sampling_rate):
|