README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: wav2vec2-base-timit-demo-google-colab
    results: []

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: 0.5090
• Wer: 0.3435

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
3.5501	1.0	500	1.9752	0.9950
0.8608	2.01	1000	0.5051	0.5035
0.43	3.01	1500	0.4485	0.4525
0.2921	4.02	2000	0.4658	0.4332
0.2248	5.02	2500	0.4262	0.4268
0.1863	6.02	3000	0.4126	0.3977
0.1542	7.03	3500	0.4795	0.3987
0.1374	8.03	4000	0.4882	0.3982
0.1231	9.04	4500	0.4312	0.3790
0.1082	10.04	5000	0.4344	0.3679
0.0949	11.04	5500	0.4720	0.3769
0.0897	12.05	6000	0.5382	0.3706
0.0816	13.05	6500	0.4946	0.3618
0.0726	14.06	7000	0.5383	0.3630
0.0656	15.06	7500	0.4944	0.3693
0.059	16.06	8000	0.5096	0.3639
0.0572	17.07	8500	0.5066	0.3572
0.0559	18.07	9000	0.5366	0.3610
0.0468	19.08	9500	0.5103	0.3604
0.0413	20.08	10000	0.5126	0.3496
0.044	21.08	10500	0.5055	0.3524
0.0351	22.09	11000	0.5526	0.3515
0.0328	23.09	11500	0.4884	0.3512
0.032	24.1	12000	0.5167	0.3474
0.0271	25.1	12500	0.5027	0.3495
0.0229	26.1	13000	0.5076	0.3444
0.0252	27.11	13500	0.5122	0.3464
0.0224	28.11	14000	0.5133	0.3447
0.0236	29.12	14500	0.5090	0.3435

Framework versions

• Transformers 4.17.0
• Pytorch 1.12.1+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1