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: 0.5353
• Wer: 0.3360

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
3.5345	1.0	500	1.8229	0.9810
0.8731	2.01	1000	0.5186	0.5165
0.4455	3.01	1500	0.4386	0.4572
0.3054	4.02	2000	0.4396	0.4286
0.2354	5.02	2500	0.4454	0.4051
0.1897	6.02	3000	0.4465	0.3925
0.1605	7.03	3500	0.4776	0.3974
0.1413	8.03	4000	0.5254	0.4062
0.1211	9.04	4500	0.5123	0.3913
0.1095	10.04	5000	0.4171	0.3711
0.1039	11.04	5500	0.4258	0.3732
0.0932	12.05	6000	0.4879	0.3701
0.0867	13.05	6500	0.4725	0.3637
0.0764	14.06	7000	0.5041	0.3636
0.0661	15.06	7500	0.4692	0.3646
0.0647	16.06	8000	0.4804	0.3612
0.0576	17.07	8500	0.5545	0.3628
0.0577	18.07	9000	0.5004	0.3557
0.0481	19.08	9500	0.5341	0.3558
0.0466	20.08	10000	0.5056	0.3514
0.0433	21.08	10500	0.4864	0.3481
0.0362	22.09	11000	0.4994	0.3473
0.0325	23.09	11500	0.5327	0.3446
0.0351	24.1	12000	0.5360	0.3445
0.0284	25.1	12500	0.5085	0.3399
0.027	26.1	13000	0.5344	0.3426
0.0247	27.11	13500	0.5310	0.3357
0.0251	28.11	14000	0.5201	0.3355
0.0228	29.12	14500	0.5353	0.3360

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1