wav2vec2-base_toy_train_data_augmented

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: 1.0238
• Wer: 0.6969

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
3.12	1.05	250	3.3998	0.9982
3.0727	2.1	500	3.1261	0.9982
1.9729	3.15	750	1.4868	0.9464
1.3213	4.2	1000	1.2598	0.8833
1.0508	5.25	1250	1.0014	0.8102
0.8483	6.3	1500	0.9475	0.7944
0.7192	7.35	1750	0.9493	0.7686
0.6447	8.4	2000	0.9872	0.7573
0.6064	9.45	2250	0.9587	0.7447
0.5384	10.5	2500	0.9332	0.7320
0.4985	11.55	2750	0.9926	0.7315
0.4643	12.6	3000	1.0008	0.7292
0.4565	13.65	3250	0.9522	0.7171
0.449	14.7	3500	0.9685	0.7140
0.4307	15.75	3750	1.0080	0.7077
0.4239	16.81	4000	0.9950	0.7023
0.389	17.86	4250	1.0260	0.7007
0.3471	18.91	4500	1.0012	0.6966
0.3276	19.96	4750	1.0238	0.6969

Framework versions

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