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: 0.5261
• Wer: 0.3351

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
3.5764	1.0	500	2.3358	1.0
0.9494	2.01	1000	0.6086	0.5448
0.4527	3.01	1500	0.4731	0.4685
0.307	4.02	2000	0.4432	0.4341
0.2366	5.02	2500	0.4343	0.4025
0.1934	6.02	3000	0.4284	0.4105
0.154	7.03	3500	0.4709	0.3936
0.14	8.03	4000	0.4296	0.3889
0.1189	9.04	4500	0.4864	0.3862
0.1057	10.04	5000	0.4903	0.3776
0.1034	11.04	5500	0.4889	0.3838
0.0899	12.05	6000	0.4680	0.3701
0.0864	13.05	6500	0.4981	0.3608
0.0714	14.06	7000	0.4608	0.3589
0.0673	15.06	7500	0.4970	0.3754
0.0606	16.06	8000	0.5344	0.3618
0.0603	17.07	8500	0.4980	0.3675
0.0588	18.07	9000	0.5339	0.3601
0.0453	19.08	9500	0.4973	0.3526
0.0433	20.08	10000	0.5359	0.3572
0.0421	21.08	10500	0.4885	0.3532
0.0359	22.09	11000	0.5184	0.3471
0.032	23.09	11500	0.5230	0.3483
0.0333	24.1	12000	0.5512	0.3474
0.0279	25.1	12500	0.5102	0.3437
0.0232	26.1	13000	0.5195	0.3384
0.0237	27.11	13500	0.5350	0.3355
0.0209	28.11	14000	0.5432	0.3368
0.023	29.12	14500	0.5261	0.3351

Framework versions

• Transformers 4.17.0
• Pytorch 1.12.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1