indic-bert-roman-urdu-binary

This model is a fine-tuned version of ai4bharat/indic-bert on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5183
• Accuracy: 0.8847
• Precision: 0.8851
• Recall: 0.8831
• F1: 0.8839

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: 32
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.6103	0.9912	56	0.5319	0.7366	0.7534	0.7445	0.7355
0.3576	2.0	113	0.3626	0.8427	0.8418	0.8428	0.8422
0.2913	2.9912	169	0.3478	0.8589	0.8582	0.8585	0.8583
0.2351	4.0	226	0.3812	0.8564	0.8755	0.8486	0.8520
0.1342	4.9912	282	0.4025	0.8652	0.8678	0.8619	0.8636
0.0733	6.0	339	0.4448	0.8639	0.8638	0.8625	0.8630
0.0325	6.9912	395	0.5974	0.8589	0.8657	0.8540	0.8565
0.0308	8.0	452	0.6238	0.8589	0.8588	0.8575	0.8580
0.01	8.9912	508	0.6391	0.8664	0.8693	0.8631	0.8649
0.0091	9.9115	560	0.6417	0.8552	0.8548	0.8540	0.8543

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0