Edit model card

he

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1138
  • Wer: 9.9943
  • Precision: 0.8917
  • Recall: 0.8913
  • F1: 0.8914
  • Precision Median: 1.0
  • Recall Median: 1.0
  • F1 Median: 1.0
  • Precision Max: 1.0
  • Recall Max: 1.0
  • F1 Max: 1.0
  • Precision Min: 0.0
  • Recall Min: 0.0
  • F1 Min: 0.0

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: 4e-05
  • 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
  • training_steps: 10000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Precision Recall F1 Precision Median Recall Median F1 Median Precision Max Recall Max F1 Max Precision Min Recall Min F1 Min
0.2168 0.04 500 0.2124 27.7691 0.6808 0.7027 0.6909 0.8125 0.8462 0.8276 1.0 1.0 1.0 0.0 0.0 0.0
0.1421 0.08 1000 0.1752 21.5191 0.7794 0.7820 0.7803 0.8889 0.8947 0.8947 1.0 1.0 1.0 0.0 0.0 0.0
0.086 0.12 1500 0.1510 17.9741 0.8044 0.8044 0.8040 0.9231 0.9231 0.9167 1.0 1.0 1.0 0.0 0.0 0.0
0.0822 0.16 2000 0.1357 17.1839 0.8070 0.8091 0.8078 0.9231 0.9231 0.9231 1.0 1.0 1.0 0.0 0.0 0.0
0.0675 0.2 2500 0.1227 14.9416 0.8324 0.8320 0.8319 0.9333 0.9333 0.9333 1.0 1.0 1.0 0.0 0.0 0.0
0.0583 0.24 3000 0.1224 14.0376 0.8528 0.8498 0.8510 0.9333 0.9333 0.9375 1.0 1.0 1.0 0.0 0.0 0.0
0.0528 0.28 3500 0.1167 13.8667 0.8393 0.8410 0.8399 0.9333 0.9333 0.9333 1.0 1.0 1.0 0.0 0.0 0.0
0.0431 0.32 4000 0.1173 13.3827 0.8546 0.8579 0.8560 0.9375 0.9412 0.9412 1.0 1.0 1.0 0.0 0.0 0.0
0.0402 0.36 4500 0.1154 12.1654 0.8695 0.8703 0.8697 0.9412 0.9412 0.9444 1.0 1.0 1.0 0.0 0.0 0.0
0.0385 0.4 5000 0.1173 11.9448 0.8593 0.8578 0.8584 0.9444 0.9444 0.9474 1.0 1.0 1.0 0.0 0.0 0.0
0.0266 0.44 5500 0.1144 12.1014 0.8706 0.8732 0.8717 0.9474 0.95 0.9583 1.0 1.0 1.0 0.0 0.0 0.0
0.021 0.48 6000 0.1161 11.7099 0.8737 0.8744 0.8739 1.0 1.0 0.9706 1.0 1.0 1.0 0.0 0.0 0.0
0.0228 0.52 6500 0.1109 10.9909 0.8685 0.8692 0.8687 1.0 1.0 0.9697 1.0 1.0 1.0 0.0 0.0 0.0
0.0172 0.56 7000 0.1075 10.7702 0.8780 0.8793 0.8784 1.0 0.9545 0.9697 1.0 1.0 1.0 0.0 0.0 0.0
0.0117 0.6 7500 0.1107 10.4356 0.8834 0.8825 0.8828 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0
0.0151 0.64 8000 0.1101 10.3146 0.8886 0.8899 0.8891 1.0 1.0 0.9744 1.0 1.0 1.0 0.0 0.0 0.0
0.0136 0.68 8500 0.1079 10.0370 0.8895 0.8903 0.8897 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0
0.0135 0.72 9000 0.1112 9.9445 0.8892 0.8892 0.8891 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0
0.0084 0.76 9500 0.1136 9.8875 0.8967 0.8964 0.8964 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0
0.0098 0.8 10000 0.1138 9.9943 0.8917 0.8913 0.8914 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0

Framework versions

  • Transformers 4.39.0.dev0
  • Pytorch 2.2.1+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0
Downloads last month
2
Safetensors
Model size
764M params
Tensor type
F32
·

Finetuned from