Edit model card

wav2vec2-bert-fon

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

  • Loss: 0.1612
  • Wer: 0.1324

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: 3e-05
  • train_batch_size: 2
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 4
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
No log 0.18 250 1.2212 0.8079
2.1756 0.35 500 0.6697 0.6058
2.1756 0.53 750 0.5137 0.4606
0.5041 0.7 1000 0.4337 0.4234
0.5041 0.88 1250 0.3452 0.3529
0.426 1.05 1500 0.2770 0.2910
0.426 1.23 1750 0.2681 0.2439
0.2916 1.4 2000 0.2423 0.2155
0.2916 1.58 2250 0.2342 0.2077
0.2591 1.75 2500 0.1986 0.1791
0.2591 1.93 2750 0.1864 0.1597
0.2261 2.1 3000 0.1712 0.1419
0.2261 2.28 3250 0.1786 0.1497
0.1564 2.45 3500 0.1612 0.1324
0.1564 2.63 3750 0.1730 0.1591
0.1542 2.8 4000 0.1558 0.1364
0.1542 2.98 4250 0.1493 0.1581
0.1559 3.15 4500 0.1489 0.1347
0.1559 3.33 4750 0.2036 0.1486
0.1992 3.5 5000 0.2644 0.1582
0.1992 3.68 5250 0.2401 0.1878
0.291 3.85 5500 0.2409 0.1749

Framework versions

  • Transformers 4.39.3
  • Pytorch 2.1.2
  • Datasets 2.18.0
  • Tokenizers 0.15.2
Downloads last month
5
Safetensors
Model size
606M params
Tensor type
F32
·

Finetuned from

Evaluation results