metadata
language:
- zh
license: apache-2.0
base_model: openai/whisper-medium
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- whucedar/datasets_stt_2
metrics:
- wer
model-index:
- name: zh-CN-model-medium-3 - whucedar
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: zh-CN
type: whucedar/datasets_stt_2
args: 'config: zh, split: test'
metrics:
- name: Wer
type: wer
value: 92.0589784096893
zh-CN-model-medium-3 - whucedar
This model is a fine-tuned version of openai/whisper-medium on the zh-CN dataset. It achieves the following results on the evaluation set:
- Loss: 0.2745
- Wer: 92.0590
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.2079 | 0.8306 | 1000 | 0.2799 | 78.5571 |
0.0942 | 1.6611 | 2000 | 0.2712 | 76.6719 |
0.0291 | 2.4917 | 3000 | 0.2717 | 85.0026 |
0.0069 | 3.3223 | 4000 | 0.2745 | 92.0590 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.3.1+cu118
- Datasets 2.20.0
- Tokenizers 0.19.1