bert-base-uncased-finetuned-math_punctuation-ignore_word_parts

This model is a fine-tuned version of bert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1981
• Precision: 0.7843
• Recall: 0.7485
• F Score: 0.7648
• Auc: 0.9248

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 12

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F Score	Auc
0.1064	0.64	500	0.1082	0.7558	0.6580	0.6964	0.9086
0.0781	1.27	1000	0.1025	0.7594	0.7226	0.7365	0.9261
0.0757	1.91	1500	0.1001	0.7945	0.6899	0.7302	0.9272
0.0538	2.54	2000	0.1061	0.7689	0.7348	0.7480	0.9298
0.0425	3.18	2500	0.1123	0.7806	0.7361	0.7560	0.9300
0.0377	3.81	3000	0.1159	0.7841	0.7437	0.7610	0.9292
0.0235	4.45	3500	0.1259	0.7786	0.7368	0.7561	0.9276
0.0227	5.08	4000	0.1436	0.7699	0.7448	0.7555	0.9277
0.0159	5.72	4500	0.1466	0.7715	0.7333	0.7514	0.9252
0.0106	6.35	5000	0.1574	0.7710	0.7456	0.7566	0.9276
0.0111	6.99	5500	0.1560	0.7694	0.7500	0.7595	0.9286
0.0074	7.62	6000	0.1645	0.7789	0.7511	0.7639	0.9305
0.0056	8.26	6500	0.1745	0.7887	0.7453	0.7648	0.9265
0.005	8.89	7000	0.1760	0.7779	0.7497	0.7629	0.9281
0.0038	9.53	7500	0.1873	0.7826	0.7505	0.7634	0.9273
0.0031	10.17	8000	0.1896	0.7855	0.7477	0.7644	0.9258
0.0026	10.8	8500	0.1929	0.7849	0.7485	0.7650	0.9263
0.0017	11.44	9000	0.1981	0.7843	0.7485	0.7648	0.9248

Framework versions

• Transformers 4.25.1
• Pytorch 2.0.0.dev20230111
• Datasets 2.8.0
• Tokenizers 0.13.2