PhishLang_PhoBERTCNN_15k

This model is a fine-tuned version of vinai/phobert-base-v2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3386
• Accuracy: 0.923
• F1: 0.9068
• Precision: 0.9180
• Recall: 0.8974

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.6996	0.5333	100	0.4548	0.8867	0.8615	0.8761	0.8503
0.6996	1.064	200	0.3524	0.9187	0.8998	0.9207	0.8844
0.6996	1.5973	300	0.3441	0.9217	0.9039	0.9223	0.8900
0.4376	2.128	400	0.3457	0.9247	0.9070	0.9296	0.8906
0.4376	2.6613	500	0.3423	0.928	0.9106	0.9366	0.8924
0.4376	3.192	600	0.3370	0.9253	0.9089	0.9245	0.8966
0.4376	3.7253	700	0.3378	0.9227	0.9074	0.9132	0.9021
0.3287	4.256	800	0.3392	0.921	0.9051	0.9126	0.8985
0.3287	4.7893	900	0.3374	0.9237	0.9078	0.9180	0.8992
0.3287	5.32	1000	0.3377	0.9223	0.9067	0.9142	0.9000
0.3287	5.8533	1100	0.3386	0.9273	0.9118	0.9248	0.9012
0.3037	6.384	1200	0.3454	0.919	0.9037	0.9064	0.9011
0.3037	6.9173	1300	0.3367	0.9233	0.9070	0.9192	0.8971
0.3037	7.448	1400	0.3373	0.9237	0.9073	0.9200	0.8970
0.291	7.9813	1500	0.3373	0.9223	0.9055	0.9196	0.8942
0.291	8.512	1600	0.3394	0.9233	0.9071	0.9189	0.8974
0.291	9.0427	1700	0.3379	0.9237	0.9075	0.9194	0.8976
0.291	9.576	1800	0.3386	0.923	0.9068	0.9180	0.8974

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0