Modelpublic

Mardiyyah/CeLLaTe-ner-3class-gaz-pubmedbert-tapt-tokenizer-adapted-wwmask-lr_2.97

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

0likes21downloads

Model Card

CeLLaTe-ner-3class-gaz-pubmedbert-tapt-tokenizer-adapted-wwmask-lr_2.97

This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-adapted-wwmask on the OTAR3088/CeLLaTe-ner-3class-iob_final dataset. It achieves the following results on the evaluation set:

—Loss: 0.0937
—Precision: 0.7856
—Recall: 0.7352
—Micro F1: 0.7596
—Weighted F1: 0.7582
—Macro F1: 0.7533
—Accuracy: 0.9818

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: 2.9725312443452025e-05
—trainbatchsize: 32
—evalbatchsize: 16
—seed: 17
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 64
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_ratio: 0.1
—num_epochs: 20
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	Micro F1	Weighted F1	Macro F1	Accuracy
0.4536	1.0	526	0.0712	0.7031	0.6562	0.6789	0.6779	0.6779	0.9779
0.0369	2.0	1052	0.0740	0.6889	0.7635	0.7243	0.7242	0.7182	0.9789
0.0178	3.0	1578	0.0800	0.7306	0.7665	0.7481	0.7492	0.7444	0.9808
0.0087	4.0	2104	0.0944	0.7856	0.7352	0.7596	0.7582	0.7533	0.9818
0.0052	5.0	2630	0.1035	0.7248	0.7850	0.7537	0.7543	0.7511	0.9805
0.0035	6.0	3156	0.1189	0.7122	0.7809	0.7450	0.7459	0.7428	0.9807
0.0025	7.0	3682	0.1351	0.6583	0.7907	0.7184	0.7248	0.7195	0.9768
0.0019	8.0	4208	0.1232	0.7549	0.7476	0.7512	0.7519	0.7486	0.9810
0.0017	9.0	4734	0.1204	0.7406	0.7691	0.7546	0.7551	0.7518	0.9809

Framework versions

—Transformers 4.48.2
—Pytorch 2.4.1+cu121
—Datasets 3.0.2
—Tokenizers 0.21.0