wav2vec2-base-en-asr-timit

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: 0.4525
• Wer: 0.3510

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: 64
• 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.6253	3.17	200	3.0613	1.0
2.9038	6.35	400	2.7513	1.0
1.5048	9.52	600	0.6193	0.5702
0.4196	12.7	800	0.4788	0.4464
0.2203	15.87	1000	0.4743	0.4098
0.1439	19.05	1200	0.4420	0.3804
0.0963	22.22	1400	0.4587	0.3620
0.073	25.4	1600	0.4681	0.3588
0.0603	28.57	1800	0.4525	0.3510

Framework versions

• Transformers 4.25.1
• Pytorch 1.12.1+cu113
• Datasets 1.18.3
• Tokenizers 0.13.2