wav2vec2-base-timit-demo-colab

This model is a fine-tuned version of facebook/wav2vec2-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4612
• Wer: 0.2963

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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.9218	4.0	500	0.6017	0.5820
0.5407	8.0	1000	0.4846	0.4388
0.2899	12.0	1500	0.4442	0.3654
0.1848	16.0	2000	0.4693	0.3396
0.1282	20.0	2500	0.4690	0.3215
0.0936	24.0	3000	0.4597	0.3125
0.0714	28.0	3500	0.4612	0.2963

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.2
• Datasets 1.18.3
• Tokenizers 0.10.3