Update app.py
Browse files
app.py
CHANGED
|
@@ -25,3 +25,32 @@ demo = gr.Interface(
|
|
| 25 |
)
|
| 26 |
|
| 27 |
demo.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
)
|
| 26 |
|
| 27 |
demo.launch()
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
| 31 |
+
import torch
|
| 32 |
+
from datasets import load_dataset
|
| 33 |
+
|
| 34 |
+
# Load model and processor
|
| 35 |
+
processor = WhisperProcessor.from_pretrained("openai/whisper-small")
|
| 36 |
+
model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-small")
|
| 37 |
+
|
| 38 |
+
# Optional: Use GPU if available
|
| 39 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 40 |
+
model = model.to(device)
|
| 41 |
+
|
| 42 |
+
# Load sample audio (here using a dummy dataset, aap apni audio file bhi use kar sakte hain)
|
| 43 |
+
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
|
| 44 |
+
sample = ds[0]["audio"]
|
| 45 |
+
|
| 46 |
+
# Prepare audio input
|
| 47 |
+
input_features = processor(sample["array"], sampling_rate=sample["sampling_rate"], return_tensors="pt").input_features
|
| 48 |
+
input_features = input_features.to(device)
|
| 49 |
+
|
| 50 |
+
# Generate transcription
|
| 51 |
+
predicted_ids = model.generate(input_features)
|
| 52 |
+
|
| 53 |
+
# Decode transcription
|
| 54 |
+
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
|
| 55 |
+
print(transcription)
|
| 56 |
+
|