xuliu15/English_32r_LoRA_1h_new

Whisper Small English 1h

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

• Loss: 0.1760

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: 0.0001
• train_batch_size: 8
• 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: 50
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.7144	1.0	39	1.4900
0.4793	2.0	78	0.5925
0.3478	3.0	117	0.5148
0.2552	4.0	156	0.4314
0.1964	5.0	195	0.1524
0.0281	6.0	234	0.1534
0.0254	7.0	273	0.1537
0.0133	8.0	312	0.1613
0.011	9.0	351	0.1601
0.0088	10.0	390	0.1619
0.007	11.0	429	0.1659
0.0067	12.0	468	0.1647
0.0056	13.0	507	0.1671
0.0053	14.0	546	0.1683
0.0047	15.0	585	0.1687
0.0041	16.0	624	0.1703
0.0037	17.0	663	0.1715
0.0037	18.0	702	0.1721
0.0035	19.0	741	0.1726
0.0034	20.0	780	0.1730
0.0032	21.0	819	0.1738
0.0031	22.0	858	0.1741
0.0033	23.0	897	0.1744
0.003	24.0	936	0.1748
0.0029	25.0	975	0.1756
0.003	26.0	1014	0.1756
0.0028	27.0	1053	0.1756
0.0027	28.0	1092	0.1760
0.0027	29.0	1131	0.1760
0.0027	30.0	1170	0.1760

Framework versions

• PEFT 0.10.0
• Transformers 4.40.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1