Redex-NL-RX-Synth-t5-small-finetuned-en-to-regex-finetuned-en-to-regex

This model is a fine-tuned version of rymaju/NL-RX-Synth-t5-small-finetuned-en-to-regex on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0811
• Semantic-accuracy: 0.912
• Gen Len: 10.71

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: 0.001
• train_batch_size: 16
• eval_batch_size: 16
• 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	Semantic-accuracy	Gen Len
0.1767	1.0	563	0.0710	0.916	10.766
0.0798	2.0	1126	0.0700	0.922	10.746
0.0677	3.0	1689	0.0711	0.92	10.732
0.0602	4.0	2252	0.0635	0.92	10.722
0.0552	5.0	2815	0.0693	0.932	10.75
0.05	6.0	3378	0.0746	0.922	10.778
0.0486	7.0	3941	0.0743	0.914	10.726
0.0462	8.0	4504	0.0712	0.902	10.772
0.0435	9.0	5067	0.0710	0.92	10.724
0.0442	10.0	5630	0.0799	0.896	10.69
0.043	11.0	6193	0.0754	0.914	10.72
0.0412	12.0	6756	0.0728	0.91	10.706
0.0395	13.0	7319	0.0755	0.914	10.71
0.0397	14.0	7882	0.0799	0.91	10.7
0.0389	15.0	8445	0.0778	0.908	10.684
0.0379	16.0	9008	0.0801	0.914	10.7
0.0369	17.0	9571	0.0806	0.912	10.708
0.0364	17.76	10000	0.0811	0.912	10.71

Framework versions

• Transformers 4.25.1
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2