---
license: apache-2.0
---

# ubitus-whisper-v1.0-medium

This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium).
The original Whisper model wasn't very accurate at translating medical terms.
We use the Whisper medium model as the basic model and use medical terminology and clinical records to fine-tune Whisper so that Whisper can correctly translate medical terminology and common consultation conversations between doctors and patients.
In addition, since Whisper does not support streaming translation mode, we have also developed an API to allow Whisper to have functions close to streaming translation to improve user experience.

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 8
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 16
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- mixed_precision_training: Native AMP


### Framework versions

- Transformers 4.39.2
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2