pic-20s_asr-scr_w2v2-base_001

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4443
• Per: 0.1499
• Pcc: 0.6371
• Ctc Loss: 0.5406
• Mse Loss: 0.8841

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: 1
• seed: 2222
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2247
• training_steps: 22470
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Pcc	Ctc Loss	Mse Loss
16.6841	3.0	2247	4.7118	0.9979	0.6160	3.7745	1.0013
4.2857	6.0	4494	4.2485	0.9979	0.6999	3.7428	0.6844
3.9118	9.0	6741	4.2032	0.9979	0.6863	3.7209	0.7501
3.5336	12.0	8988	3.8740	0.9976	0.6645	3.1591	0.9697
2.1131	15.0	11235	2.0043	0.2726	0.6564	1.1426	0.8936
0.9858	18.0	13482	1.6048	0.1817	0.6377	0.7083	0.8783
0.7106	21.0	15729	1.5797	0.1625	0.6447	0.6061	0.9394
0.5928	24.0	17976	1.4856	0.1552	0.6392	0.5624	0.8977
0.525	27.0	20223	1.4673	0.1515	0.6343	0.5471	0.8972
0.4862	30.0	22470	1.4443	0.1499	0.6371	0.5406	0.8841

Framework versions

• Transformers 4.38.1
• Pytorch 2.0.1
• Datasets 2.16.1
• Tokenizers 0.15.2