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: 0.5452
• Wer: 0.3296

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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5557	1.0	500	1.9362	1.0072
0.867	2.01	1000	0.5197	0.5173
0.4281	3.01	1500	0.4609	0.4552
0.3002	4.02	2000	0.4066	0.4129
0.2252	5.02	2500	0.4122	0.3952
0.1857	6.02	3000	0.4650	0.3990
0.1541	7.03	3500	0.4784	0.3834
0.1372	8.03	4000	0.3875	0.3805
0.1213	9.04	4500	0.5606	0.4002
0.1043	10.04	5000	0.4713	0.3762
0.0972	11.04	5500	0.4770	0.3692
0.0876	12.05	6000	0.4755	0.3671
0.0812	13.05	6500	0.4854	0.3616
0.0705	14.06	7000	0.4380	0.3659
0.0759	15.06	7500	0.5025	0.3516
0.0709	16.06	8000	0.5310	0.3577
0.0572	17.07	8500	0.5097	0.3561
0.0572	18.07	9000	0.5150	0.3510
0.0482	19.08	9500	0.4954	0.3488
0.0703	20.08	10000	0.5279	0.3512
0.0457	21.08	10500	0.5336	0.3459
0.036	22.09	11000	0.5471	0.3440
0.0368	23.09	11500	0.5109	0.3417
0.0342	24.1	12000	0.5506	0.3415
0.0318	25.1	12500	0.5291	0.3357
0.03	26.1	13000	0.5347	0.3363
0.026	27.11	13500	0.5475	0.3318
0.0232	28.11	14000	0.5628	0.3332
0.0246	29.12	14500	0.5452	0.3296

Framework versions

• Transformers 4.17.0
• Pytorch 1.12.1+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1