wav2vec2-base-timit-demo-colab

This model is a fine-tuned version of facebook/wav2vec2-base on the timit_asr dataset. It achieves the following results on the evaluation set:

• Loss: 0.4243
• Wer: 0.2830

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
3.5112	3.45	500	1.1699	0.8236
0.5349	6.9	1000	0.3911	0.3609
0.1875	10.34	1500	0.3993	0.3170
0.1113	13.79	2000	0.3870	0.3046
0.0778	17.24	2500	0.4056	0.2963
0.0561	20.69	3000	0.3781	0.2918
0.0461	24.14	3500	0.4186	0.2857
0.0375	27.59	4000	0.4243	0.2830

Framework versions

• Transformers 4.21.1
• Pytorch 1.12.0
• Datasets 2.4.0
• Tokenizers 0.12.1