Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,37 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- eu
|
| 4 |
+
license: mit
|
| 5 |
+
tags:
|
| 6 |
+
- automatic-speech-recognition
|
| 7 |
+
- whisper
|
| 8 |
+
- hf-asr-leaderboard
|
| 9 |
+
- audio
|
| 10 |
+
- euskara
|
| 11 |
+
- basque
|
| 12 |
+
metrics:
|
| 13 |
+
- wer
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# Whisper Base Eus (Euskara)
|
| 17 |
+
|
| 18 |
+
Este es un modelo fine-tuneado basado en [openai/whisper-base](https://huggingface.co/openai/whisper-base) dise帽ado espec铆ficamente para la **Transcripci贸n Autom谩tica de Voz (ASR)** en Euskera (Vasco).
|
| 19 |
+
|
| 20 |
+
## 馃殌 Uso del modelo
|
| 21 |
+
|
| 22 |
+
Para usar el modelo directamente en Python a trav茅s de la librer铆a `transformers`, necesitas tener instalado `transformers` y `torch` (junto a `ffmpeg` en tu sistema si vas a procesar archivos de audio).
|
| 23 |
+
|
| 24 |
+
```python
|
| 25 |
+
from transformers import pipeline
|
| 26 |
+
|
| 27 |
+
# Cargar el pipeline de transcripci贸n
|
| 28 |
+
transcriber = pipeline(
|
| 29 |
+
"automatic-speech-recognition",
|
| 30 |
+
model="ikerua/whisper-base-eus",
|
| 31 |
+
chunk_length_s=30
|
| 32 |
+
)
|
| 33 |
+
|
| 34 |
+
# Transcribir un archivo de audio (.wav, .mp3, etc.)
|
| 35 |
+
resultado = transcriber("ruta_a_tu_audio.mp3", generate_kwargs={"language": "basque", "task": "transcribe"})
|
| 36 |
+
|
| 37 |
+
print(resultado["text"])
|