spellcorrector_20_02_050_qwerty_v7

This model is a fine-tuned version of Buseak/spellcorrector_18_02_050_qwerty_v6 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0005
• Precision: 0.9995
• Recall: 1.0
• F1: 0.9997
• Accuracy: 0.9999

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
0.0674	1.0	967	0.0375	0.9849	0.9802	0.9825	0.9894
0.0481	2.0	1934	0.0259	0.9903	0.9871	0.9887	0.9924
0.0387	3.0	2901	0.0210	0.9909	0.9877	0.9893	0.9938
0.032	4.0	3868	0.0161	0.9957	0.9920	0.9938	0.9952
0.0277	5.0	4835	0.0127	0.9952	0.9941	0.9946	0.9962
0.0238	6.0	5802	0.0105	0.9973	0.9952	0.9962	0.9970
0.0208	7.0	6769	0.0080	0.9973	0.9952	0.9962	0.9977
0.0176	8.0	7736	0.0064	0.9962	0.9952	0.9957	0.9981
0.0154	9.0	8703	0.0049	0.9979	0.9968	0.9973	0.9986
0.0134	10.0	9670	0.0039	0.9957	0.9968	0.9962	0.9989
0.0114	11.0	10637	0.0030	0.9989	0.9995	0.9992	0.9992
0.0105	12.0	11604	0.0022	0.9995	0.9995	0.9995	0.9994
0.0089	13.0	12571	0.0018	0.9995	1.0	0.9997	0.9996
0.0079	14.0	13538	0.0013	0.9995	0.9995	0.9995	0.9997
0.0067	15.0	14505	0.0010	0.9995	0.9995	0.9995	0.9998
0.0056	16.0	15472	0.0009	0.9995	1.0	0.9997	0.9998
0.0051	17.0	16439	0.0007	0.9995	1.0	0.9997	0.9999
0.0047	18.0	17406	0.0006	0.9995	1.0	0.9997	0.9999
0.0046	19.0	18373	0.0005	0.9995	1.0	0.9997	0.9999
0.0041	20.0	19340	0.0005	0.9995	1.0	0.9997	0.9999

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2