---
license: cc-by-nc-4.0
pipeline_tag: automatic-speech-recognition
base_model:
- openai/whisper-small
library_name: transformers
language:
- ami
- trv
- bnn
- pwn
- tay
- tsu
- tao
- dru
- xsy
- pyu
- szy
- ckv
- sxr
- ssf
- xnb
---
# Model Card for whisper-small-formosan-all

<!-- Provide a quick summary of what the model is/does. -->
This model is a fine-tuned version of the Taiwanese indigenous [openai/whisper-small](https://huggingface.co/openai/whisper-small).   
Note: we use indonesian as whisper language id

### Training process
The training of the model was performed with the following hyperparameters
- Batch size: 32*4 (on 4 L40s GPU)
- Gradient accumulation steps: 8
- Total steps: 1600
- Learning rate: 1.25e-5
- Data augmentation: No
- Optimizer: schedule_free_adamw
- LR scheduler type: constant


### How to use

```python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model_id = "formospeech/whisper-small-formosan-all"
model = AutoModelForSpeechSeq2Seq.from_pretrained(
    model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
)
model.to(device)
processor = AutoProcessor.from_pretrained(model_id)
pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    max_new_tokens=128,
    chunk_length_s=30,
    batch_size=16,
    torch_dtype=torch_dtype,
    device=device,
)
generate_kwargs = {"language": "id"}
transcription = pipe("path/to/my_audio.wav", generate_kwargs=generate_kwargs)
print(transcription)
```