Update app.py
Browse files
app.py
CHANGED
|
@@ -14,7 +14,7 @@ app = FastAPI()
|
|
| 14 |
device = 0 if torch.cuda.is_available() else -1
|
| 15 |
|
| 16 |
# Load Whisper model and processor
|
| 17 |
-
model_name = "openai/whisper-
|
| 18 |
model = WhisperForConditionalGeneration.from_pretrained(model_name)
|
| 19 |
processor = WhisperProcessor.from_pretrained(model_name)
|
| 20 |
|
|
@@ -22,11 +22,12 @@ processor = WhisperProcessor.from_pretrained(model_name)
|
|
| 22 |
forced_decoder_ids = processor.get_decoder_prompt_ids(language="portuguese", task="transcribe")
|
| 23 |
model.config.forced_decoder_ids = forced_decoder_ids
|
| 24 |
|
| 25 |
-
# Initialize the ASR pipeline with the modified model
|
| 26 |
asr_pipeline = pipeline(
|
| 27 |
"automatic-speech-recognition",
|
| 28 |
model=model,
|
| 29 |
-
|
|
|
|
| 30 |
device=device
|
| 31 |
)
|
| 32 |
|
|
|
|
| 14 |
device = 0 if torch.cuda.is_available() else -1
|
| 15 |
|
| 16 |
# Load Whisper model and processor
|
| 17 |
+
model_name = "openai/whisper-large-v2" # You can change to other variants like "openai/whisper-small"
|
| 18 |
model = WhisperForConditionalGeneration.from_pretrained(model_name)
|
| 19 |
processor = WhisperProcessor.from_pretrained(model_name)
|
| 20 |
|
|
|
|
| 22 |
forced_decoder_ids = processor.get_decoder_prompt_ids(language="portuguese", task="transcribe")
|
| 23 |
model.config.forced_decoder_ids = forced_decoder_ids
|
| 24 |
|
| 25 |
+
# Initialize the ASR pipeline with the modified model and processor
|
| 26 |
asr_pipeline = pipeline(
|
| 27 |
"automatic-speech-recognition",
|
| 28 |
model=model,
|
| 29 |
+
tokenizer=processor.tokenizer, # Explicitly set the tokenizer from the processor
|
| 30 |
+
feature_extractor=processor.feature_extractor, # Also set the feature extractor
|
| 31 |
device=device
|
| 32 |
)
|
| 33 |
|