Edit model card

wav2vec2-base-finetuned-ks

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: 0.5766
  • Accuracy: 0.8308

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: 1.5e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 15

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 1.0 7 0.7247 0.7462
No log 2.0 14 0.6844 0.7615
0.4279 3.0 21 0.7254 0.7462
0.4279 4.0 28 0.5891 0.8
0.4279 5.0 35 0.6991 0.7462
0.4478 6.0 42 0.6579 0.7615
0.4478 7.0 49 0.6164 0.8
0.4478 8.0 56 0.6191 0.8077
0.4194 9.0 63 0.5766 0.8308
0.4194 10.0 70 0.5704 0.8154
0.4194 11.0 77 0.6518 0.8
0.3833 12.0 84 0.6190 0.8077
0.3833 13.0 91 0.5693 0.8231
0.3833 14.0 98 0.5628 0.8231
0.3607 15.0 105 0.5741 0.8154

Framework versions

  • Transformers 4.11.3
  • Pytorch 1.10.0+cu111
  • Datasets 2.0.0
  • Tokenizers 0.10.3
Downloads last month
3