bartpho-vietnews-sum

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

• Loss: 3.3348
• Rouge1: 0.5983
• Rouge2: 0.2976
• Rougel: 0.4114
• Rougelsum: 0.4114

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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: 3
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
3.6491	0.3828	2000	3.5148	0.5802	0.2748	0.3923	0.3922
3.3865	0.7656	4000	3.4149	0.5888	0.2845	0.4010	0.4009
3.0592	1.1483	6000	3.3750	0.5923	0.2911	0.4055	0.4054
2.9393	1.5311	8000	3.3433	0.5915	0.2912	0.4067	0.4066
2.8404	1.9139	10000	3.3122	0.5952	0.2938	0.4082	0.4080
2.6108	2.2967	12000	3.3402	0.5964	0.2948	0.4086	0.4085
2.5911	2.6795	14000	3.3348	0.5983	0.2976	0.4114	0.4114

Framework versions

• Transformers 4.52.1
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1