phobert-base_v2

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

• Loss: 0.3362
• Accuracy: 0.9482
• Precision Macro: 0.8854
• Recall Macro: 0.8318
• F1 Macro: 0.8543
• F1 Weighted: 0.9464

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
0.4592	1.0	90	0.2280	0.9356	0.8885	0.7440	0.7800	0.9283
0.1801	2.0	180	0.1823	0.9476	0.8617	0.8443	0.8523	0.9469
0.1221	3.0	270	0.1834	0.9482	0.8795	0.8359	0.8548	0.9467
0.1071	4.0	360	0.1868	0.9520	0.9086	0.8096	0.8447	0.9486
0.0817	5.0	450	0.2031	0.9526	0.8980	0.8393	0.8635	0.9508
0.065	6.0	540	0.2240	0.9501	0.8908	0.8084	0.8389	0.9469
0.0574	7.0	630	0.2219	0.9501	0.8625	0.8701	0.8662	0.9504
0.0481	8.0	720	0.2503	0.9469	0.8752	0.8266	0.8472	0.9451
0.0362	9.0	810	0.2489	0.9495	0.8822	0.8121	0.8392	0.9466
0.0319	10.0	900	0.2584	0.9501	0.8784	0.8413	0.8577	0.9488
0.0263	11.0	990	0.2774	0.9488	0.8800	0.8281	0.8498	0.9469
0.0199	12.0	1080	0.2790	0.9501	0.8780	0.8416	0.8577	0.9488
0.0114	13.0	1170	0.2955	0.9476	0.8733	0.8393	0.8546	0.9463
0.0126	14.0	1260	0.3105	0.9501	0.8953	0.8331	0.8586	0.9481
0.0125	15.0	1350	0.3147	0.9482	0.8773	0.8397	0.8564	0.9469
0.0106	16.0	1440	0.3247	0.9469	0.8861	0.8350	0.8567	0.9453
0.0065	17.0	1530	0.3419	0.9476	0.8751	0.8274	0.8476	0.9458
0.0072	18.0	1620	0.3406	0.9469	0.8933	0.8185	0.8475	0.9444
0.0058	19.0	1710	0.3389	0.9495	0.8904	0.8328	0.8566	0.9476
0.0064	20.0	1800	0.3362	0.9482	0.8854	0.8318	0.8543	0.9464

Framework versions

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