Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -7,6 +7,9 @@ import torch
|
|
7 |
|
8 |
from transformers import pipeline
|
9 |
|
|
|
|
|
|
|
10 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
11 |
model_repo_id = "emlinking/wav2vec2-large-xls-r-300m-tsm-asr-v6"
|
12 |
|
@@ -15,23 +18,25 @@ if torch.cuda.is_available():
|
|
15 |
else:
|
16 |
torch_dtype = torch.float32
|
17 |
|
18 |
-
|
|
|
19 |
|
20 |
# @spaces.GPU #[uncomment to use ZeroGPU]
|
21 |
def infer(
|
22 |
audio,
|
23 |
target
|
24 |
):
|
|
|
25 |
sampling_rate, wav = audio
|
26 |
if wav.ndim > 1:
|
27 |
wav = wav.mean(axis=1)
|
28 |
wav = wav.astype(np.float32)
|
29 |
wav /= np.max(np.abs(wav))
|
30 |
-
|
|
|
31 |
|
32 |
# compare texts
|
33 |
-
|
34 |
-
d_toks = [(i[2:], i[0] if i[0] != " " else None) for i in d.compare(target, user_pron)]
|
35 |
return (user_pron, d_toks)
|
36 |
|
37 |
css = """
|
@@ -52,7 +57,7 @@ with gr.Blocks(css=css) as demo:
|
|
52 |
label='Comparison',
|
53 |
combine_adjacent=True,
|
54 |
show_legend=True,
|
55 |
-
color_map=
|
56 |
)
|
57 |
input_audio.input(fn=infer, inputs=[input_audio, target], outputs=[output, diff])
|
58 |
|
|
|
7 |
|
8 |
from transformers import pipeline
|
9 |
|
10 |
+
# ################ CHANGE THIS TO CHANGE THE LANGUAGE ###################### #
|
11 |
+
from TaiwaneseHokkien import TaiwaneseHokkien
|
12 |
+
|
13 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
14 |
model_repo_id = "emlinking/wav2vec2-large-xls-r-300m-tsm-asr-v6"
|
15 |
|
|
|
18 |
else:
|
19 |
torch_dtype = torch.float32
|
20 |
|
21 |
+
language = TaiwaneseHokkien(device=device, torch_dtype=torch_dtype)
|
22 |
+
# ########################################################################## #
|
23 |
|
24 |
# @spaces.GPU #[uncomment to use ZeroGPU]
|
25 |
def infer(
|
26 |
audio,
|
27 |
target
|
28 |
):
|
29 |
+
# preprocess
|
30 |
sampling_rate, wav = audio
|
31 |
if wav.ndim > 1:
|
32 |
wav = wav.mean(axis=1)
|
33 |
wav = wav.astype(np.float32)
|
34 |
wav /= np.max(np.abs(wav))
|
35 |
+
|
36 |
+
user_pron = language.asr(wav)
|
37 |
|
38 |
# compare texts
|
39 |
+
d_toks = language.compare(target, user_pron)
|
|
|
40 |
return (user_pron, d_toks)
|
41 |
|
42 |
css = """
|
|
|
57 |
label='Comparison',
|
58 |
combine_adjacent=True,
|
59 |
show_legend=True,
|
60 |
+
color_map=language.compare_colors
|
61 |
)
|
62 |
input_audio.input(fn=infer, inputs=[input_audio, target], outputs=[output, diff])
|
63 |
|