Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
|
3 |
import torch
|
|
|
4 |
|
5 |
# Load the pre-trained model and processor
|
6 |
model_name = "facebook/s2t-wav2vec2-large-en-ar"
|
@@ -9,13 +10,20 @@ processor = Wav2Vec2Processor.from_pretrained(model_name)
|
|
9 |
|
10 |
# Define a function for the ASR model
|
11 |
def transcribe(audio):
|
|
|
|
|
|
|
|
|
12 |
# Process the audio
|
13 |
inputs = processor(audio, return_tensors="pt", sampling_rate=16000)
|
|
|
14 |
# Get the model's predictions
|
15 |
logits = model(input_values=inputs.input_values).logits
|
|
|
16 |
# Decode the predicted text
|
17 |
predicted_ids = logits.argmax(dim=-1)
|
18 |
transcription = processor.decode(predicted_ids[0])
|
|
|
19 |
return transcription
|
20 |
|
21 |
# Define the Gradio interface
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
|
3 |
import torch
|
4 |
+
import numpy as np
|
5 |
|
6 |
# Load the pre-trained model and processor
|
7 |
model_name = "facebook/s2t-wav2vec2-large-en-ar"
|
|
|
10 |
|
11 |
# Define a function for the ASR model
|
12 |
def transcribe(audio):
|
13 |
+
# Convert the audio into a format compatible with the processor
|
14 |
+
if isinstance(audio, np.ndarray):
|
15 |
+
audio = audio.flatten() # Ensure it's a 1D array
|
16 |
+
|
17 |
# Process the audio
|
18 |
inputs = processor(audio, return_tensors="pt", sampling_rate=16000)
|
19 |
+
|
20 |
# Get the model's predictions
|
21 |
logits = model(input_values=inputs.input_values).logits
|
22 |
+
|
23 |
# Decode the predicted text
|
24 |
predicted_ids = logits.argmax(dim=-1)
|
25 |
transcription = processor.decode(predicted_ids[0])
|
26 |
+
|
27 |
return transcription
|
28 |
|
29 |
# Define the Gradio interface
|