Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,6 @@ import torch
|
|
5 |
import torchaudio
|
6 |
|
7 |
|
8 |
-
df= pd.read_csv('native_words_subset.csv')
|
9 |
|
10 |
torch._C._jit_override_can_fuse_on_cpu(False)
|
11 |
torch._C._jit_override_can_fuse_on_gpu(False)
|
@@ -13,47 +12,14 @@ torch._C._jit_set_texpr_fuser_enabled(False)
|
|
13 |
torch._C._jit_set_nvfuser_enabled(False)
|
14 |
|
15 |
loader = torch.jit.load("audio_loader.pt")
|
16 |
-
model = torch.jit.load('
|
17 |
|
18 |
vocab = model.text_transform.vocab.itos
|
19 |
vocab[-1] = ''
|
20 |
|
21 |
-
def convert_probs(probs):
|
22 |
-
ids = probs.argmax(1)[0]
|
23 |
-
s = []
|
24 |
-
if vocab[ids[0]]: s.append(vocab[ids[0]])
|
25 |
-
for i in range(1,len(ids)):
|
26 |
-
if ids[i-1] != ids[i]:
|
27 |
-
new = vocab[ids[i]]
|
28 |
-
if new: s.append(new)
|
29 |
-
#return '.'.join(s)
|
30 |
-
return s
|
31 |
-
|
32 |
|
33 |
def predict(path):
|
34 |
audio = loader(path)
|
35 |
-
|
36 |
-
return convert_probs(probs)
|
37 |
-
|
38 |
-
|
39 |
-
from difflib import SequenceMatcher
|
40 |
|
41 |
-
|
42 |
-
return SequenceMatcher(None, a, b).ratio()
|
43 |
-
|
44 |
-
def compare(chosen_word, path):
|
45 |
-
etalons = [list(val.split('.')) for val in df.loc[df['replica'] == chosen_word, 'transcription'].values]
|
46 |
-
user = predict(path)
|
47 |
-
coeff = 0.0
|
48 |
-
idx=0
|
49 |
-
for i in range(len(etalons)):
|
50 |
-
new_coeff = similar(user, etalons[i])
|
51 |
-
if new_coeff > coeff:
|
52 |
-
coeff = new_coeff
|
53 |
-
idx=i
|
54 |
-
return f'The similarity coefficient of your pronunciation and the pronunciation of a native speaker is {coeff}. The closer the coefficient is to 1, the better.' + '\nYour pronunciation: [' + ''.join(user) + ']\nClosest native pronunciation: [' + ''.join(etalons[idx]) + ']'
|
55 |
-
|
56 |
-
|
57 |
-
word_choice = gr.inputs.Dropdown(sorted(list(df['replica'].unique())), label="Choose a word")
|
58 |
-
|
59 |
-
gr.Interface(fn=compare, inputs=[word_choice, gr.inputs.Audio(source='microphone', type='filepath', optional=True)], outputs= 'text').launch(debug=True)
|
|
|
5 |
import torchaudio
|
6 |
|
7 |
|
|
|
8 |
|
9 |
torch._C._jit_override_can_fuse_on_cpu(False)
|
10 |
torch._C._jit_override_can_fuse_on_gpu(False)
|
|
|
12 |
torch._C._jit_set_nvfuser_enabled(False)
|
13 |
|
14 |
loader = torch.jit.load("audio_loader.pt")
|
15 |
+
model = torch.jit.load('QuartzNet15x5Base_En_1.pt').eval()
|
16 |
|
17 |
vocab = model.text_transform.vocab.itos
|
18 |
vocab[-1] = ''
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
def predict(path):
|
22 |
audio = loader(path)
|
23 |
+
return model.predict(audio)
|
|
|
|
|
|
|
|
|
24 |
|
25 |
+
gr.Interface(fn=predict, predict=[gr.inputs.Audio(source='microphone', type='filepath'], outputs= 'text').launch(debug=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|