xuliu15/Frisian_32r_LoRA_1h

LoRA-Frisian-1h

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_6_1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8286

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: 0.001
• 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: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.6913	1.0	89	1.1389
0.6757	2.0	178	0.9073
0.3686	3.0	267	0.8561
0.2062	4.0	356	0.8493
0.11	5.0	445	0.8448
0.0652	6.0	534	0.8446
0.0393	7.0	623	0.8239
0.0205	8.0	712	0.8243
0.0137	9.0	801	0.8279
0.0095	10.0	890	0.8105
0.0065	11.0	979	0.8192
0.0076	12.0	1068	0.8240
0.006	13.0	1157	0.8255
0.0048	14.0	1246	0.8272
0.0038	15.0	1335	0.8286

Framework versions

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