mirzaaa10/banglabert_finetuned_v2

results

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

• Loss: 0.7260
• Accuracy: 0.7417
• F1: 0.7267

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.9409	1.0	35	0.9145	0.6583	0.6004
0.7658	2.0	70	0.7490	0.725	0.6680
0.6112	3.0	105	0.6922	0.7333	0.6772
0.4237	4.0	140	0.6434	0.7583	0.7283
0.327	5.0	175	0.6385	0.7667	0.7513
0.2258	6.0	210	0.6861	0.7333	0.7266
0.1018	7.0	245	0.7522	0.7583	0.7327
0.0794	8.0	280	0.8856	0.7667	0.7506
0.0543	9.0	315	0.9427	0.7417	0.7368
0.03	10.0	350	0.9346	0.75	0.7339
0.0252	11.0	385	1.0011	0.7333	0.7274
0.0324	12.0	420	1.0499	0.75	0.7366
0.0139	13.0	455	1.0554	0.7333	0.7173
0.014	14.0	490	1.0881	0.7583	0.7492
0.0131	15.0	525	1.0949	0.7583	0.7492

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1