indic-bert-roman-urdu-fine-grained

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: 0.8501
• Accuracy: 0.7678
• Precision: 0.6945
• Recall: 0.6537
• F1: 0.6720

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: 2
• total_train_batch_size: 64
• 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
1.1237	1.0	113	1.0947	0.5342	0.1068	0.2	0.1393
0.9606	2.0	226	0.8776	0.6689	0.4456	0.3188	0.2779
0.7784	3.0	339	0.6443	0.7896	0.7017	0.6830	0.6899
0.5626	4.0	452	0.5167	0.8302	0.7561	0.7371	0.7422
0.5613	5.0	565	0.4285	0.8634	0.7931	0.7849	0.7850
0.4232	6.0	678	0.3543	0.8867	0.8295	0.8072	0.8155
0.3376	7.0	791	0.2546	0.9293	0.8850	0.8757	0.8802
0.2759	8.0	904	0.2079	0.9469	0.9085	0.9132	0.9103
0.2029	9.0	1017	0.1564	0.9606	0.9370	0.9276	0.9322
0.137	10.0	1130	0.1364	0.9685	0.9558	0.9399	0.9477

Framework versions

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