README.md

metadata

library_name: transformers
language:
  - lg
license: mit
base_model: facebook/w2v-bert-2.0
tags:
  - generated_from_trainer
datasets:
  - yogera
metrics:
  - wer
model-index:
  - name: wav2vec2-bert
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Yogera
          type: yogera
        metrics:
          - name: Wer
            type: wer
            value: 0.14867316851893853

wav2vec2-bert

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the Yogera dataset. It achieves the following results on the evaluation set:

• Loss: 0.2216
• Wer: 0.1487
• Cer: 0.0334

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.6681	1.0	235	0.2226	0.2616	0.0533
0.1666	2.0	470	0.1639	0.2013	0.0410
0.1249	3.0	705	0.1608	0.1912	0.0416
0.101	4.0	940	0.1573	0.1835	0.0416
0.0833	5.0	1175	0.1567	0.1697	0.0378
0.0715	6.0	1410	0.1589	0.1564	0.0346
0.0624	7.0	1645	0.1634	0.1728	0.0408
0.0541	8.0	1880	0.1592	0.1559	0.0341
0.0464	9.0	2115	0.1788	0.1546	0.0336
0.0434	10.0	2350	0.1641	0.1575	0.0353
0.0385	11.0	2585	0.1854	0.1498	0.0333
0.0358	12.0	2820	0.1915	0.1504	0.0345
0.0308	13.0	3055	0.1747	0.1514	0.0328
0.0283	14.0	3290	0.1966	0.1449	0.0329
0.0274	15.0	3525	0.1882	0.1535	0.0342
0.0246	16.0	3760	0.2199	0.1588	0.0362
0.0212	17.0	3995	0.2108	0.1572	0.0355
0.0188	18.0	4230	0.2173	0.1453	0.0320
0.017	19.0	4465	0.2358	0.1444	0.0324
0.0177	20.0	4700	0.2280	0.1548	0.0339
0.0174	21.0	4935	0.2142	0.1484	0.0322
0.0138	22.0	5170	0.2315	0.1489	0.0338
0.0122	23.0	5405	0.2116	0.1483	0.0341
0.0125	24.0	5640	0.2216	0.1487	0.0334

Framework versions

• Transformers 4.45.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.1