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: 0.5707
• Wer: 0.3388

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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5072	1.0	500	1.8786	0.9741
0.8836	2.01	1000	0.5147	0.5317
0.4576	3.01	1500	0.4774	0.4591
0.3056	4.02	2000	0.4393	0.4343
0.2349	5.02	2500	0.4404	0.4022
0.1946	6.02	3000	0.4564	0.3991
0.1624	7.03	3500	0.4428	0.3947
0.1421	8.03	4000	0.4312	0.3878
0.131	9.04	4500	0.4345	0.3853
0.1115	10.04	5000	0.4318	0.3753
0.1024	11.04	5500	0.5053	0.3798
0.0895	12.05	6000	0.5044	0.3782
0.0856	13.05	6500	0.4893	0.3665
0.0755	14.06	7000	0.4868	0.3662
0.0724	15.06	7500	0.5084	0.3681
0.0635	16.06	8000	0.5367	0.3530
0.0603	17.07	8500	0.5255	0.3604
0.0609	18.07	9000	0.5407	0.3678
0.0486	19.08	9500	0.5312	0.3630
0.047	20.08	10000	0.5498	0.3518
0.0437	21.08	10500	0.5326	0.3571
0.0379	22.09	11000	0.5644	0.3608
0.035	23.09	11500	0.5956	0.3539
0.0333	24.1	12000	0.5967	0.3517
0.0289	25.1	12500	0.5274	0.3399
0.0268	26.1	13000	0.5609	0.3406
0.0256	27.11	13500	0.5451	0.3448
0.0249	28.11	14000	0.5804	0.3413
0.0236	29.12	14500	0.5707	0.3388

Framework versions

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