whisper-judicial-ct2
Modelo CTranslate2 del fine-tune judicial en español para Faster-Whisper.
Origen
- Base:
openai/whisper-large-v3-turbo - Fine-tune: LoRA r=32, alpha=64
- Conversión:
ct2-transformers-converter --quantization float16
Uso
from faster_whisper import WhisperModel
model = WhisperModel("92kdd0/whisper-judicial-ct2", device="cpu", compute_type="int8")
segments, _ = model.transcribe("audio.wav", language="es")
En TRANSCRIPTORJUDICIAL: DEFAULT_MODEL_SIZE=92kdd0/whisper-judicial-ct2
Métricas
| Métrica | Valor |
|---|---|
| WER normalizado (eval set adaptador) | 5.90 % |
| WER/CER/MER normalizados (audio prueba judicial) | 0.00 % |
- Downloads last month
- 22
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for 92kdd0/whisper-judicial-ct2
Base model
openai/whisper-large-v3 Finetuned
openai/whisper-large-v3-turbo