wav2vec2-16

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: 3.1016
• Wer: 1.0

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: 0.001
• train_batch_size: 32
• 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: 400
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.6682	1.37	200	3.3138	1.0
2.8751	2.74	400	2.9984	1.0
2.8697	4.11	600	3.0827	1.0
2.866	5.48	800	3.0697	1.0
2.8655	6.85	1000	3.1083	1.0
2.8629	8.22	1200	3.0888	1.0
2.8651	9.59	1400	3.2852	1.0
2.8601	10.96	1600	3.1155	1.0
2.8617	12.33	1800	3.1958	1.0
2.8595	13.7	2000	3.1070	1.0
2.858	15.07	2200	3.1483	1.0
2.8564	16.44	2400	3.0906	1.0
2.8561	17.81	2600	3.1412	1.0
2.8574	19.18	2800	3.0783	1.0
2.8543	20.55	3000	3.0624	1.0
2.8549	21.92	3200	3.0914	1.0
2.8556	23.29	3400	3.0735	1.0
2.8557	24.66	3600	3.1791	1.0
2.8576	26.03	3800	3.0645	1.0
2.8528	27.4	4000	3.1190	1.0
2.8551	28.77	4200	3.1016	1.0

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1