wav2vec2-tcrs-runtest

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: 3.1370
• 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: 0.0001
• train_batch_size: 2
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
22.437	1.43	10	36.3252	1.0
14.7939	2.86	20	10.7441	1.0
4.1824	4.29	30	3.7354	1.0
3.289	5.71	40	3.5265	1.0
3.1639	7.14	50	3.2868	1.0
3.1107	8.57	60	3.3268	1.0
3.0737	10.0	70	3.1149	1.0
3.0273	11.43	80	3.2031	1.0
3.0422	12.86	90	3.0771	1.0
2.9957	14.29	100	3.0418	1.0
2.9894	15.71	110	3.0321	1.0
2.9997	17.14	120	3.0545	1.0
2.9806	18.57	130	2.9936	1.0
2.969	20.0	140	3.0322	1.0
2.9692	21.43	150	3.0238	1.0
2.9638	22.86	160	3.0407	1.0
2.969	24.29	170	3.2487	1.0
2.9783	25.71	180	3.1248	1.0
2.9576	27.14	190	3.0880	1.0
2.968	28.57	200	3.0962	1.0
2.9784	30.0	210	3.1370	1.0

Framework versions

• Transformers 4.11.3
• Pytorch 1.11.0+cu102
• Datasets 1.18.3
• Tokenizers 0.10.3