wav2vec2-base-librispeech32

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: 0.3137
• Wer: 0.1101

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: 32
• 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
3.6899	2.07	500	2.7071	0.9991
0.5624	4.13	1000	0.4322	0.2304
0.1855	6.2	1500	0.4234	0.2023
0.1224	8.26	2000	0.4044	0.1852
0.0928	10.33	2500	0.4644	0.2213
0.0766	12.4	3000	0.3669	0.1459
0.0655	14.46	3500	0.3215	0.1414
0.0544	16.53	4000	0.3524	0.1292
0.0475	18.6	4500	0.4299	0.1818
0.0405	20.66	5000	0.3026	0.1226
0.0361	22.73	5500	0.3132	0.1206
0.0329	24.79	6000	0.3409	0.1086
0.03	26.86	6500	0.3183	0.1099
0.0276	28.93	7000	0.3137	0.1101

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3