wav2vec2-base-timit-demo-google-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: 0.4634
• Wer: 0.3367

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: 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: 1000
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.6019	1.0	500	2.4586	1.0
0.9594	2.01	1000	0.5023	0.5122
0.4324	3.01	1500	0.4808	0.4703
0.2991	4.02	2000	0.4098	0.4208
0.2257	5.02	2500	0.4883	0.4264
0.18	6.02	3000	0.4441	0.3914
0.1524	7.03	3500	0.4360	0.3869
0.1315	8.03	4000	0.4448	0.3783
0.1101	9.04	4500	0.4570	0.3704
0.1017	10.04	5000	0.4252	0.3680
0.0863	11.04	5500	0.4492	0.3606
0.0798	12.05	6000	0.4241	0.3604
0.0688	13.05	6500	0.4585	0.3535
0.0608	14.06	7000	0.4491	0.3488
0.0524	15.06	7500	0.4550	0.3456
0.0502	16.06	8000	0.4570	0.3453
0.0458	17.07	8500	0.4680	0.3421
0.0395	18.07	9000	0.4663	0.3390
0.0352	19.08	9500	0.4634	0.3367

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.2
• Datasets 1.18.3
• Tokenizers 0.15.2