Modelpublic

Abhi964/Paraphrase_indicBERT_onfull_FT3

sourceHugging Facemitupdated 2y agoView on Hugging Face

0likes12downloads

Model Card

ParaphraseindicBERTonfull_FT3

This model is a fine-tuned version of ai4bharat/indic-bert on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 1.0320
—Accuracy: 0.789
—F1: 0.7885

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: 1.4638638566821256e-05
—trainbatchsize: 32
—evalbatchsize: 64
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 14

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.5707	1.0	157	0.5842	0.6895	0.6630
0.4831	2.0	314	0.5444	0.7435	0.7420
0.4363	3.0	471	0.4700	0.775	0.7730
0.3548	4.0	628	0.4781	0.7765	0.7763
0.2468	5.0	785	0.5416	0.786	0.7858
0.2046	6.0	942	0.6293	0.7775	0.7768
0.127	7.0	1099	0.6558	0.7815	0.7802
0.1042	8.0	1256	0.9524	0.742	0.7381
0.0653	9.0	1413	1.0619	0.7485	0.7450
0.0253	10.0	1570	1.0320	0.789	0.7885
0.0405	11.0	1727	1.1028	0.7795	0.7794
0.0106	12.0	1884	1.1150	0.784	0.7840
0.0098	13.0	2041	1.1362	0.785	0.7850
0.0331	14.0	2198	1.1453	0.785	0.7850

Framework versions

—Transformers 4.49.0
—Pytorch 2.6.0+cu124
—Datasets 3.3.2
—Tokenizers 0.21.0