spellcorrector_18_02_050_qwerty_v6

This model is a fine-tuned version of Buseak/spellcorrector_17_02_050_qwerty on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0028
• Precision: 0.9968
• Recall: 0.9941
• F1: 0.9954
• Accuracy: 0.9993

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
0.0887	1.0	967	0.0551	0.9876	0.9801	0.9838	0.9842
0.0684	2.0	1934	0.0415	0.9930	0.9844	0.9887	0.9881
0.0581	3.0	2901	0.0343	0.9924	0.9855	0.9890	0.9899
0.0487	4.0	3868	0.0280	0.9925	0.9882	0.9903	0.9917
0.0425	5.0	4835	0.0241	0.9930	0.9882	0.9906	0.9930
0.0382	6.0	5802	0.0209	0.9946	0.9882	0.9914	0.9940
0.0333	7.0	6769	0.0168	0.9951	0.9909	0.9930	0.9950
0.0294	8.0	7736	0.0148	0.9941	0.9909	0.9925	0.9957
0.0265	9.0	8703	0.0121	0.9946	0.9909	0.9927	0.9964
0.0238	10.0	9670	0.0103	0.9952	0.9919	0.9935	0.9970
0.0216	11.0	10637	0.0090	0.9978	0.9930	0.9954	0.9974
0.0193	12.0	11604	0.0076	0.9952	0.9930	0.9941	0.9979
0.0175	13.0	12571	0.0065	0.9973	0.9936	0.9954	0.9982
0.016	14.0	13538	0.0055	0.9973	0.9936	0.9954	0.9985
0.0137	15.0	14505	0.0045	0.9968	0.9936	0.9952	0.9988
0.0127	16.0	15472	0.0039	0.9973	0.9941	0.9957	0.9990
0.0118	17.0	16439	0.0034	0.9978	0.9941	0.9960	0.9991
0.0111	18.0	17406	0.0030	0.9968	0.9941	0.9954	0.9992
0.0104	19.0	18373	0.0029	0.9968	0.9941	0.9954	0.9993
0.0099	20.0	19340	0.0028	0.9968	0.9941	0.9954	0.9993

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2