wav2vec2-base_toy_train_data_fast_10pct

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: 1.3087
• Wer: 0.7175

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
3.1309	1.05	250	3.4541	0.9982
3.0499	2.1	500	3.0231	0.9982
1.4839	3.15	750	1.4387	0.9257
1.1697	4.2	1000	1.3729	0.8792
0.9353	5.25	1250	1.2608	0.8445
0.7298	6.3	1500	1.1867	0.8052
0.6418	7.35	1750	1.2414	0.7997
0.5698	8.4	2000	1.2240	0.7766
0.5084	9.45	2250	1.1910	0.7687
0.4912	10.5	2500	1.2241	0.7617
0.4144	11.55	2750	1.2412	0.7477
0.4153	12.6	3000	1.2736	0.7511
0.405	13.65	3250	1.2827	0.7328
0.3852	14.7	3500	1.1981	0.7331
0.3829	15.75	3750	1.3035	0.7347
0.3538	16.81	4000	1.3003	0.7240
0.3385	17.86	4250	1.3354	0.7304
0.3108	18.91	4500	1.2983	0.7229
0.3037	19.96	4750	1.3087	0.7175

Framework versions

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