luyotw/test-medium-XieLongJie-11-36

Fine-tune 資訊

• 原始模型: openai/whisper-medium
• 使用音訊數量: 5044
• 使用音訊總長: 2.87 小時
• 音訊平均長度: 2.05 秒
• GPU: NVIDIA H100 PCIe x 1
• 訓練時間: 01:22:01
• 模型大小: 2.85 GB
• 訓練參數:
  • batch size: 23
  • eval batch size: 12
  • gradient checkpointing: False
  • fp16: False
  • bf16: True

---

Fine-tuned Whisper model for Legislative Yuan of Taiwan

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0396
• Wer: 83.7458

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: 23
• eval_batch_size: 12
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0034	4.5455	1000	0.0319	85.2174
0.0001	9.0909	2000	0.0360	83.2107
0.0	13.6364	3000	0.0382	83.2776
0.0	18.1818	4000	0.0392	83.4114
0.0	22.7273	5000	0.0396	83.7458

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1
• Datasets 3.5.0
• Tokenizers 0.21.1