wav2vec2-large-xlsr-53_toy_train_data_augment_0.1

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4658
• Wer: 0.5037

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.447	1.05	250	3.3799	1.0
3.089	2.1	500	3.4868	1.0
3.063	3.15	750	3.3155	1.0
2.4008	4.2	1000	1.2934	0.8919
1.618	5.25	1250	0.7847	0.7338
1.3038	6.3	1500	0.6459	0.6712
1.2074	7.35	1750	0.5705	0.6269
1.1062	8.4	2000	0.5267	0.5843
1.026	9.45	2250	0.5108	0.5683
0.9505	10.5	2500	0.5066	0.5568
0.893	11.55	2750	0.5161	0.5532
0.8535	12.6	3000	0.4994	0.5341
0.8462	13.65	3250	0.4626	0.5262
0.8334	14.7	3500	0.4593	0.5197
0.842	15.75	3750	0.4651	0.5126
0.7678	16.81	4000	0.4687	0.5120
0.7873	17.86	4250	0.4716	0.5070
0.7486	18.91	4500	0.4657	0.5033
0.7073	19.96	4750	0.4658	0.5037

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu102
• Datasets 2.0.0
• Tokenizers 0.11.6