wav2vec2-base-timit-demo-google-colab

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: 3.0392
• 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: 8
• 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: 150

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.2993	8.0	200	3.0327	1.0
3.0806	16.0	400	3.0476	1.0
3.0219	24.0	600	3.0472	1.0
3.0179	32.0	800	3.0435	1.0
3.0157	40.0	1000	3.0546	1.0
3.0146	48.0	1200	3.0484	1.0
3.0139	56.0	1400	3.0344	1.0
3.0118	64.0	1600	3.0351	1.0
3.0114	72.0	1800	3.0559	1.0
3.0114	80.0	2000	3.0526	1.0
3.0108	88.0	2200	3.0417	1.0
3.0092	96.0	2400	3.0629	1.0
3.0089	104.0	2600	3.0352	1.0
3.0083	112.0	2800	3.0503	1.0
3.0078	120.0	3000	3.0529	1.0
3.0072	128.0	3200	3.0378	1.0
3.0068	136.0	3400	3.0481	1.0
3.0063	144.0	3600	3.0392	1.0

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.3
• Tokenizers 0.13.3