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: 0.5501
• Wer: 0.3424

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
3.5448	1.0	500	2.5044	1.0
1.0167	2.01	1000	0.5435	0.5278
0.4453	3.01	1500	0.4450	0.4534
0.3	4.02	2000	0.4401	0.4245
0.2304	5.02	2500	0.4146	0.4022
0.1889	6.02	3000	0.4241	0.3927
0.1573	7.03	3500	0.4545	0.3878
0.1363	8.03	4000	0.4936	0.3940
0.1213	9.04	4500	0.4964	0.3806
0.108	10.04	5000	0.4931	0.3826
0.0982	11.04	5500	0.5373	0.3778
0.0883	12.05	6000	0.4978	0.3733
0.0835	13.05	6500	0.5189	0.3728
0.0748	14.06	7000	0.4608	0.3692
0.068	15.06	7500	0.4827	0.3608
0.0596	16.06	8000	0.5022	0.3661
0.056	17.07	8500	0.5482	0.3646
0.0565	18.07	9000	0.5158	0.3573
0.0487	19.08	9500	0.4910	0.3513
0.0444	20.08	10000	0.5771	0.3580
0.045	21.08	10500	0.5160	0.3539
0.0363	22.09	11000	0.5367	0.3503
0.0313	23.09	11500	0.5773	0.3500
0.0329	24.1	12000	0.5683	0.3508
0.0297	25.1	12500	0.5355	0.3464
0.0272	26.1	13000	0.5317	0.3450
0.0256	27.11	13500	0.5602	0.3443
0.0242	28.11	14000	0.5586	0.3419
0.0239	29.12	14500	0.5501	0.3424

Framework versions

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