JunWorks/whisperSmall_LoRA_zhTW

Whisper Small LoRA tuned zh-TW

This model is a fine-tuned version of openai/whisper-small on the Common Voice 13.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2094
• CER: 13.1583%

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: 8
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.2284	1.0	1453	0.2305
0.2221	2.0	2906	0.2183
0.1939	3.0	4359	0.2149
0.2085	4.0	5812	0.2125
0.2202	5.0	7265	0.2112
0.2097	6.0	8718	0.2103
0.2072	7.0	10171	0.2095
0.1906	8.0	11624	0.2094
0.1818	9.0	13077	0.2091
0.2054	10.0	14530	0.2093
0.1597	11.0	15983	0.2094
0.1797	12.0	17436	0.2095
0.204	13.0	18889	0.2096
0.1872	14.0	20342	0.2094
0.1879	15.0	21795	0.2094

Framework versions

• PEFT 0.9.1.dev0
• Transformers 4.40.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2