phobert-large_v2

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

• Loss: 0.3603
• Accuracy: 0.9520
• Precision Macro: 0.8966
• Recall Macro: 0.8344
• F1 Macro: 0.8599
• F1 Weighted: 0.9500

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: 3e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Use 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision Macro	Recall Macro	F1 Macro	F1 Weighted
0.4609	1.0	90	0.2266	0.9387	0.9153	0.7341	0.7705	0.9298
0.1892	2.0	180	0.2063	0.9419	0.8750	0.8032	0.8303	0.9390
0.1302	3.0	270	0.1810	0.9551	0.8937	0.8696	0.8809	0.9545
0.0972	4.0	360	0.1825	0.9533	0.9077	0.8396	0.8671	0.9513
0.0676	5.0	450	0.2135	0.9514	0.8737	0.8546	0.8636	0.9507
0.0567	6.0	540	0.2387	0.9507	0.9005	0.8334	0.8605	0.9487
0.047	7.0	630	0.2487	0.9419	0.8520	0.8401	0.8457	0.9414
0.0405	8.0	720	0.3009	0.9501	0.9101	0.8004	0.8370	0.9462
0.0225	9.0	810	0.2780	0.9514	0.8963	0.8217	0.8506	0.9488
0.0275	10.0	900	0.2952	0.9514	0.8945	0.8421	0.8643	0.9498
0.0141	11.0	990	0.3188	0.9488	0.8690	0.8611	0.8649	0.9486
0.0122	12.0	1080	0.3221	0.9520	0.8983	0.8261	0.8545	0.9496
0.0091	13.0	1170	0.3291	0.9526	0.9042	0.8430	0.8684	0.9509
0.0053	14.0	1260	0.3365	0.9476	0.8795	0.8270	0.8490	0.9456
0.0065	15.0	1350	0.3530	0.9520	0.9009	0.8345	0.8613	0.9500
0.0038	16.0	1440	0.3478	0.9520	0.9033	0.8261	0.8561	0.9495
0.0031	17.0	1530	0.3586	0.9501	0.9001	0.8329	0.8601	0.9481
0.0016	18.0	1620	0.3596	0.9514	0.8947	0.8421	0.8645	0.9498
0.0015	19.0	1710	0.3594	0.9520	0.8966	0.8344	0.8599	0.9500
0.0014	20.0	1800	0.3603	0.9520	0.8966	0.8344	0.8599	0.9500

Framework versions

• Transformers 4.55.0
• Pytorch 2.7.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4