model_phoneme_onSet3

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

  • eval_loss: 0.1119
  • eval_0_precision: 1.0
  • eval_0_recall: 0.9355
  • eval_0_f1-score: 0.9667
  • eval_0_support: 31
  • eval_1_precision: 0.9231
  • eval_1_recall: 1.0
  • eval_1_f1-score: 0.9600
  • eval_1_support: 24
  • eval_2_precision: 1.0
  • eval_2_recall: 0.9545
  • eval_2_f1-score: 0.9767
  • eval_2_support: 22
  • eval_3_precision: 0.9524
  • eval_3_recall: 1.0
  • eval_3_f1-score: 0.9756
  • eval_3_support: 20
  • eval_accuracy: 0.9691
  • eval_macro avg_precision: 0.9689
  • eval_macro avg_recall: 0.9725
  • eval_macro avg_f1-score: 0.9698
  • eval_macro avg_support: 97
  • eval_weighted avg_precision: 0.9711
  • eval_weighted avg_recall: 0.9691
  • eval_weighted avg_f1-score: 0.9691
  • eval_weighted avg_support: 97
  • eval_wer: 0.0986
  • eval_mtrix: [[0, 1, 2, 3], [0, 29, 1, 0, 1], [1, 0, 24, 0, 0], [2, 0, 1, 21, 0], [3, 0, 0, 0, 20]]
  • eval_runtime: 5.7078
  • eval_samples_per_second: 16.994
  • eval_steps_per_second: 2.278
  • step: 0

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: 0.0003
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 200
  • num_epochs: 70
  • mixed_precision_training: Native AMP

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Datasets 2.8.0
  • Tokenizers 0.13.2
Downloads last month
7
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.