Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,16 +1,36 @@
|
|
|
|
1 |
from transformers import pipeline
|
2 |
import gradio as gr
|
|
|
3 |
|
4 |
-
|
|
|
|
|
|
|
|
|
5 |
|
6 |
|
|
|
7 |
def transcribe(audio):
|
8 |
-
text =
|
9 |
return text
|
10 |
|
11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
iface = gr.Interface(
|
13 |
-
fn=
|
14 |
inputs=gr.Audio(sources="microphone", type="filepath"),
|
15 |
outputs="text",
|
16 |
title="Whisper Bangla",
|
|
|
1 |
+
import spaces
|
2 |
from transformers import pipeline
|
3 |
import gradio as gr
|
4 |
+
import torch
|
5 |
|
6 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
7 |
+
asr = pipeline(model="asif00/whisper-bangla").to(device=device)
|
8 |
+
ser = pipeline("text2text-generation", model="asif00/mbart_bn_error_correction").to(
|
9 |
+
device=device
|
10 |
+
)
|
11 |
|
12 |
|
13 |
+
@spaces.GPU
|
14 |
def transcribe(audio):
|
15 |
+
text = asr(audio)["text"]
|
16 |
return text
|
17 |
|
18 |
|
19 |
+
@spaces.GPU
|
20 |
+
def correction(text):
|
21 |
+
corrected_text = ser(text)
|
22 |
+
print(corrected_text)
|
23 |
+
return corrected_text
|
24 |
+
|
25 |
+
|
26 |
+
def transcribe_and_correct(audio):
|
27 |
+
text = transcribe(audio)
|
28 |
+
corrected_text = correction(text)
|
29 |
+
return corrected_text
|
30 |
+
|
31 |
+
|
32 |
iface = gr.Interface(
|
33 |
+
fn=transcribe_and_correct,
|
34 |
inputs=gr.Audio(sources="microphone", type="filepath"),
|
35 |
outputs="text",
|
36 |
title="Whisper Bangla",
|