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: 0.5112
• Wer: 0.9988

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
3.5557	1.0	500	1.6786	1.0
0.8407	2.01	1000	0.5356	0.9988
0.4297	3.01	1500	0.4431	0.9988
0.2989	4.02	2000	0.4191	0.9988
0.2338	5.02	2500	0.4251	0.9988
0.1993	6.02	3000	0.4618	0.9988
0.1585	7.03	3500	0.4577	0.9988
0.1386	8.03	4000	0.4099	0.9982
0.1234	9.04	4500	0.4945	0.9988
0.1162	10.04	5000	0.4597	0.9988
0.1008	11.04	5500	0.4563	0.9988
0.0894	12.05	6000	0.5157	0.9988
0.083	13.05	6500	0.5027	0.9988
0.0735	14.06	7000	0.4905	0.9994
0.0686	15.06	7500	0.4552	0.9988
0.0632	16.06	8000	0.5522	0.9988
0.061	17.07	8500	0.4874	0.9988
0.0626	18.07	9000	0.5243	0.9988
0.0475	19.08	9500	0.4798	0.9988
0.0447	20.08	10000	0.5250	0.9988
0.0432	21.08	10500	0.5195	0.9988
0.0358	22.09	11000	0.5008	0.9988
0.0319	23.09	11500	0.5376	0.9988
0.0334	24.1	12000	0.5149	0.9988
0.0269	25.1	12500	0.4911	0.9988
0.0275	26.1	13000	0.4907	0.9988
0.027	27.11	13500	0.4992	0.9988
0.0239	28.11	14000	0.5021	0.9988
0.0233	29.12	14500	0.5112	0.9988

Framework versions

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