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: 0.5093
• Wer: 0.3413

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
3.5009	1.0	500	1.6207	0.9471
0.8414	2.01	1000	0.5128	0.5033
0.4366	3.01	1500	0.4449	0.4450
0.3015	4.02	2000	0.3835	0.4108
0.2281	5.02	2500	0.3989	0.4109
0.1914	6.02	3000	0.4286	0.3982
0.1555	7.03	3500	0.4547	0.3889
0.1349	8.03	4000	0.3876	0.3779
0.1252	9.04	4500	0.4460	0.3810
0.1066	10.04	5000	0.3905	0.3772
0.0979	11.04	5500	0.4469	0.3646
0.0883	12.05	6000	0.4547	0.3612
0.0801	13.05	6500	0.4741	0.3645
0.0709	14.06	7000	0.4682	0.3592
0.0665	15.06	7500	0.4689	0.3647
0.0579	16.06	8000	0.5330	0.3622
0.0556	17.07	8500	0.4885	0.3575
0.0547	18.07	9000	0.4936	0.3543
0.0462	19.08	9500	0.4928	0.3524
0.0475	20.08	10000	0.5286	0.3525
0.0426	21.08	10500	0.5100	0.3550
0.0364	22.09	11000	0.5372	0.3493
0.0306	23.09	11500	0.5049	0.3443
0.0314	24.1	12000	0.5223	0.3519
0.0261	25.1	12500	0.5380	0.3486
0.0257	26.1	13000	0.5326	0.3484
0.0252	27.11	13500	0.5299	0.3446
0.0226	28.11	14000	0.5174	0.3424
0.0232	29.12	14500	0.5093	0.3413

Framework versions

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