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: 0.5253
• Wer: 0.3406

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
3.4884	1.0	500	1.6139	1.0293
0.8373	2.01	1000	0.5286	0.5266
0.4394	3.01	1500	0.4933	0.4678
0.2974	4.02	2000	0.4159	0.4268
0.2268	5.02	2500	0.4288	0.4074
0.1901	6.02	3000	0.4407	0.3852
0.1627	7.03	3500	0.4599	0.3849
0.1397	8.03	4000	0.4330	0.3803
0.1342	9.04	4500	0.4661	0.3785
0.1165	10.04	5000	0.4518	0.3745
0.1	11.04	5500	0.4714	0.3899
0.0881	12.05	6000	0.4985	0.3848
0.0794	13.05	6500	0.5074	0.3672
0.0707	14.06	7000	0.5692	0.3681
0.0669	15.06	7500	0.4722	0.3814
0.0589	16.06	8000	0.5738	0.3784
0.0562	17.07	8500	0.5183	0.3696
0.0578	18.07	9000	0.5473	0.3841
0.0473	19.08	9500	0.4918	0.3655
0.0411	20.08	10000	0.5258	0.3517
0.0419	21.08	10500	0.5256	0.3501
0.0348	22.09	11000	0.5511	0.3597
0.0328	23.09	11500	0.5054	0.3560
0.0314	24.1	12000	0.5327	0.3537
0.0296	25.1	12500	0.5142	0.3446
0.0251	26.1	13000	0.5155	0.3411
0.0249	27.11	13500	0.5344	0.3414
0.0225	28.11	14000	0.5193	0.3408
0.0226	29.12	14500	0.5253	0.3406

Framework versions

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