wav2vec2_tun_1.0

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3582
• Wer: 0.2211
• Cer: 0.0770

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: 16
• 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: 500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
5.9699	3.7234	700	2.8579	1.0	0.9915
1.5582	7.4468	1400	0.4742	0.3345	0.1187
0.5216	11.1702	2100	0.3712	0.2768	0.0975
0.3359	14.8936	2800	0.3522	0.2484	0.0867
0.2514	18.6170	3500	0.3611	0.2403	0.0833
0.217	22.3404	4200	0.3557	0.2266	0.0787
0.1873	26.0638	4900	0.3569	0.2245	0.0786
0.183	29.7872	5600	0.3582	0.2211	0.0770

Framework versions

• Transformers 4.41.1
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1