wav2vec2-large-xlsr-53_toy_train_data_augmented

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: 0.5016
• Wer: 0.4656

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
3.418	1.05	250	3.4171	1.0
3.0886	2.1	500	3.4681	1.0
2.9422	3.15	750	2.6151	1.0
1.3195	4.2	1000	0.8789	0.7739
0.9154	5.25	1250	0.6364	0.6518
0.6519	6.3	1500	0.5682	0.5949
0.5622	7.35	1750	0.5273	0.5625
0.4965	8.4	2000	0.4891	0.5283
0.4283	9.45	2250	0.5018	0.5260
0.4019	10.5	2500	0.5016	0.5006
0.3585	11.55	2750	0.5047	0.5003
0.3275	12.6	3000	0.5148	0.4866
0.3427	13.65	3250	0.5035	0.4786
0.3229	14.7	3500	0.4855	0.4768
0.3332	15.75	3750	0.5040	0.4769
0.2861	16.81	4000	0.5138	0.4669
0.3029	17.86	4250	0.5133	0.4670
0.2633	18.91	4500	0.5063	0.4637
0.2621	19.96	4750	0.5016	0.4656

Framework versions

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