Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -14,6 +14,7 @@ model = WhisperForConditionalGeneration.from_pretrained(
|
|
14 |
import torch
|
15 |
|
16 |
processor = WhisperProcessor.from_pretrained('WajeehAzeemX/whisper-smal-ar-testing-kale-5000')
|
|
|
17 |
|
18 |
from transformers import GenerationConfig, WhisperForConditionalGeneration
|
19 |
generation_config = GenerationConfig.from_pretrained("openai/whisper-small") # if you are using a multilingual model
|
@@ -43,7 +44,7 @@ async def transcribe_audio(request: Request):
|
|
43 |
input_features = processor(audio_array, sampling_rate=sampling_rate, return_tensors="pt").input_features
|
44 |
|
45 |
# Generate token ids
|
46 |
-
predicted_ids = model.generate(input_features,
|
47 |
|
48 |
# Decode token ids to text
|
49 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
|
|
|
14 |
import torch
|
15 |
|
16 |
processor = WhisperProcessor.from_pretrained('WajeehAzeemX/whisper-smal-ar-testing-kale-5000')
|
17 |
+
forced_decoder_ids = processor.get_decoder_prompt_ids(language="arabic", task="transcribe")
|
18 |
|
19 |
from transformers import GenerationConfig, WhisperForConditionalGeneration
|
20 |
generation_config = GenerationConfig.from_pretrained("openai/whisper-small") # if you are using a multilingual model
|
|
|
44 |
input_features = processor(audio_array, sampling_rate=sampling_rate, return_tensors="pt").input_features
|
45 |
|
46 |
# Generate token ids
|
47 |
+
predicted_ids = model.generate(input_features, forced_decoder_ids=forced_decoder_ids)
|
48 |
|
49 |
# Decode token ids to text
|
50 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
|