MahaPhrase_IndicBERT_Finetune_2

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: 2.1777
• Accuracy: 0.792
• F1: 0.7907

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
No log	1.0	282	0.5638	0.732	0.7293
0.5176	2.0	564	0.6016	0.732	0.7320
0.5176	3.0	846	0.6152	0.752	0.752
0.4173	4.0	1128	0.9357	0.732	0.7288
0.4173	5.0	1410	0.6378	0.744	0.7333
0.4395	6.0	1692	0.6688	0.768	0.7668
0.4395	7.0	1974	0.9093	0.764	0.7620
0.321	8.0	2256	0.9201	0.748	0.7468
0.2227	9.0	2538	0.7661	0.796	0.7909
0.2227	10.0	2820	1.3013	0.776	0.7613
0.1222	11.0	3102	1.2891	0.788	0.7870
0.1222	12.0	3384	1.3665	0.828	0.8257
0.0602	13.0	3666	1.4113	0.832	0.8307
0.0602	14.0	3948	1.8551	0.784	0.784
0.0222	15.0	4230	1.6908	0.788	0.7842
0.0134	16.0	4512	1.8649	0.788	0.7868
0.0134	17.0	4794	1.8400	0.796	0.7951
0.0022	18.0	5076	1.7347	0.8	0.7995
0.0022	19.0	5358	1.7748	0.796	0.7932
0.0039	20.0	5640	2.0701	0.808	0.8072
0.0039	21.0	5922	2.1207	0.808	0.8072
0.0008	22.0	6204	2.1592	0.8	0.7990
0.0008	23.0	6486	2.1735	0.796	0.7948
0.0	24.0	6768	2.1772	0.796	0.7948
0.0	25.0	7050	2.1777	0.792	0.7907

Framework versions

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