ascend_with_timit

This model is a fine-tuned version of GleamEyeBeast/ascend_with_timit on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8013
• Wer: 0.4781
• Cer: 0.1727

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.4026	1.0	890	1.3419	0.9083	0.3670
1.1926	2.0	1780	0.9730	0.6491	0.2585
0.9104	3.0	2670	0.8483	0.5368	0.1963
0.7718	4.0	3560	0.8122	0.4913	0.1791
0.7013	5.0	4450	0.8013	0.4781	0.1727

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.0+cu111
• Datasets 2.0.0
• Tokenizers 0.11.6