Edit model card

spellcorrector_20_02_050_qwerty_v13

This model is a fine-tuned version of Buseak/spellcorrector_20_02_050_qwerty_v11 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0000
  • Precision: 1.0
  • Recall: 0.9995
  • F1: 0.9997
  • Accuracy: 1.0000

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: 2e-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
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
0.0398 1.0 967 0.0161 0.9941 0.9898 0.9919 0.9953
0.0227 2.0 1934 0.0103 0.9952 0.9919 0.9935 0.9970
0.017 3.0 2901 0.0072 0.9936 0.9957 0.9946 0.9978
0.0126 4.0 3868 0.0048 0.9978 0.9946 0.9962 0.9986
0.0104 5.0 4835 0.0034 0.9989 0.9973 0.9981 0.9990
0.0082 6.0 5802 0.0022 0.9989 0.9979 0.9984 0.9993
0.0064 7.0 6769 0.0015 0.9989 0.9973 0.9981 0.9996
0.0055 8.0 7736 0.0013 0.9995 0.9989 0.9992 0.9996
0.0042 9.0 8703 0.0007 0.9989 0.9979 0.9984 0.9998
0.0035 10.0 9670 0.0004 0.9995 0.9989 0.9992 0.9999
0.0032 11.0 10637 0.0003 0.9995 0.9989 0.9992 0.9999
0.0025 12.0 11604 0.0002 0.9995 0.9989 0.9992 1.0000
0.0022 13.0 12571 0.0002 0.9995 0.9989 0.9992 1.0000
0.002 14.0 13538 0.0001 1.0 0.9989 0.9995 1.0000
0.0015 15.0 14505 0.0001 0.9995 1.0 0.9997 1.0000
0.0013 16.0 15472 0.0001 0.9995 1.0 0.9997 1.0000
0.0012 17.0 16439 0.0001 1.0 1.0 1.0 1.0000
0.001 18.0 17406 0.0000 1.0 0.9995 0.9997 1.0000
0.0009 19.0 18373 0.0000 1.0 0.9995 0.9997 1.0000
0.0009 20.0 19340 0.0000 1.0 0.9995 0.9997 1.0000

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.17.1
  • Tokenizers 0.15.2
Downloads last month
621
Safetensors
Model size
132M params
Tensor type
F32
·

Finetuned from