Harveenchadha
commited on
Commit
•
5b6d82c
1
Parent(s):
07d59d1
Refactored Code
Browse files
app.py
CHANGED
@@ -2,97 +2,64 @@ import soundfile as sf
|
|
2 |
import torch
|
3 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor,Wav2Vec2ProcessorWithLM
|
4 |
import gradio as gr
|
5 |
-
import scipy.signal as sps
|
6 |
import sox
|
7 |
import subprocess
|
8 |
|
9 |
-
def convert(inputfile, outfile):
|
10 |
-
sox_tfm = sox.Transformer()
|
11 |
-
sox_tfm.set_output_format(
|
12 |
-
file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
|
13 |
-
)
|
14 |
-
#print(this is not done)
|
15 |
-
sox_tfm.build(inputfile, outfile)
|
16 |
|
17 |
-
def
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
|
|
|
|
23 |
|
24 |
|
25 |
def resampler(input_file_path, output_file_path):
|
26 |
-
#output_file_path = output_folder_path + input_file_path.split('/')[-1]
|
27 |
-
|
28 |
command = (
|
29 |
f"ffmpeg -hide_banner -loglevel panic -i {input_file_path} -ar 16000 -ac 1 -bits_per_raw_sample 16 -vn "
|
30 |
f"{output_file_path}"
|
31 |
)
|
32 |
subprocess.call(command, shell=True)
|
33 |
|
34 |
-
def parse_transcription_with_lm(wav_file):
|
35 |
-
input_values = read_file_and_process(wav_file)
|
36 |
-
|
37 |
-
# with torch.no_grad():
|
38 |
-
# logits = model(**input_values).logits[0].cpu().numpy()
|
39 |
-
# print(logits)
|
40 |
-
# int_result = processor_with_LM.decode(logits = logits, output_word_offsets=False,
|
41 |
-
# beam_width=128
|
42 |
-
# )
|
43 |
-
# print(int_result)
|
44 |
-
# transcription = int_result.text.replace('<s>','')
|
45 |
-
|
46 |
-
|
47 |
-
with torch.no_grad():
|
48 |
-
logits = model(**input_values).logits
|
49 |
|
|
|
50 |
result = processor_with_LM.batch_decode(logits.cpu().numpy())
|
51 |
text = result.text
|
52 |
transcription = text[0].replace('<s>','')
|
53 |
return transcription
|
54 |
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
speech, _ = sf.read(filename + "16k.wav")
|
60 |
-
inputs = processor(speech, sampling_rate=16_000, return_tensors="pt", padding=True)
|
61 |
-
|
62 |
-
return inputs
|
63 |
|
64 |
def parse(wav_file, applyLM):
|
|
|
|
|
|
|
|
|
65 |
if applyLM:
|
66 |
return parse_transcription_with_lm(wav_file)
|
67 |
else:
|
68 |
return parse_transcription(wav_file)
|
69 |
|
70 |
-
def parse_transcription(wav_file):
|
71 |
-
input_values = read_file_and_process(wav_file)
|
72 |
-
with torch.no_grad():
|
73 |
-
logits = model(**input_values).logits
|
74 |
-
#logits = model(input_values).logits
|
75 |
-
predicted_ids = torch.argmax(logits, dim=-1)
|
76 |
-
|
77 |
-
transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
|
78 |
-
return transcription
|
79 |
|
80 |
model_id = "Harveenchadha/vakyansh-wav2vec2-hindi-him-4200"
|
81 |
processor = Wav2Vec2Processor.from_pretrained(model_id)
|
82 |
processor_with_LM = Wav2Vec2ProcessorWithLM.from_pretrained(model_id)
|
83 |
model = Wav2Vec2ForCTC.from_pretrained(model_id)
|
84 |
-
|
85 |
|
86 |
|
87 |
input_ = gr.Audio(source="microphone", type="filepath")
|
88 |
-
#input_ = gr.inputs.Audio(source="microphone", type="numpy")
|
89 |
txtbox = gr.Textbox(
|
90 |
label="Output from model will appear here:",
|
91 |
lines=5
|
92 |
)
|
93 |
-
|
94 |
chkbox = gr.Checkbox(label="Apply LM", value=False)
|
95 |
|
|
|
96 |
gr.Interface(parse, inputs = [input_, chkbox], outputs=txtbox,
|
97 |
streaming=True, interactive=True,
|
98 |
analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);
|
|
|
2 |
import torch
|
3 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor,Wav2Vec2ProcessorWithLM
|
4 |
import gradio as gr
|
|
|
5 |
import sox
|
6 |
import subprocess
|
7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
|
9 |
+
def read_file_and_process(wav_file):
|
10 |
+
filename = wav_file.split('.')[0]
|
11 |
+
filename_16k = filename + "16k.wav"
|
12 |
+
resampler(wav_file, filename_16k)
|
13 |
+
speech, _ = sf.read(filename_16k)
|
14 |
+
inputs = processor(speech, sampling_rate=16_000, return_tensors="pt", padding=True)
|
15 |
+
|
16 |
+
return inputs
|
17 |
|
18 |
|
19 |
def resampler(input_file_path, output_file_path):
|
|
|
|
|
20 |
command = (
|
21 |
f"ffmpeg -hide_banner -loglevel panic -i {input_file_path} -ar 16000 -ac 1 -bits_per_raw_sample 16 -vn "
|
22 |
f"{output_file_path}"
|
23 |
)
|
24 |
subprocess.call(command, shell=True)
|
25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
+
def parse_transcription_with_lm(logits):
|
28 |
result = processor_with_LM.batch_decode(logits.cpu().numpy())
|
29 |
text = result.text
|
30 |
transcription = text[0].replace('<s>','')
|
31 |
return transcription
|
32 |
|
33 |
+
def parse_transcription(logits):
|
34 |
+
predicted_ids = torch.argmax(logits, dim=-1)
|
35 |
+
transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
|
36 |
+
return transcription
|
|
|
|
|
|
|
|
|
37 |
|
38 |
def parse(wav_file, applyLM):
|
39 |
+
input_values = read_file_and_process(wav_file)
|
40 |
+
with torch.no_grad():
|
41 |
+
logits = model(**input_values).logits
|
42 |
+
|
43 |
if applyLM:
|
44 |
return parse_transcription_with_lm(wav_file)
|
45 |
else:
|
46 |
return parse_transcription(wav_file)
|
47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
model_id = "Harveenchadha/vakyansh-wav2vec2-hindi-him-4200"
|
50 |
processor = Wav2Vec2Processor.from_pretrained(model_id)
|
51 |
processor_with_LM = Wav2Vec2ProcessorWithLM.from_pretrained(model_id)
|
52 |
model = Wav2Vec2ForCTC.from_pretrained(model_id)
|
|
|
53 |
|
54 |
|
55 |
input_ = gr.Audio(source="microphone", type="filepath")
|
|
|
56 |
txtbox = gr.Textbox(
|
57 |
label="Output from model will appear here:",
|
58 |
lines=5
|
59 |
)
|
|
|
60 |
chkbox = gr.Checkbox(label="Apply LM", value=False)
|
61 |
|
62 |
+
|
63 |
gr.Interface(parse, inputs = [input_, chkbox], outputs=txtbox,
|
64 |
streaming=True, interactive=True,
|
65 |
analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);
|