bart_lora_finetune

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

• Loss: 1.8519
• Rouge1: 0.422
• Rouge2: 0.193
• Rougel: 0.2864
• Rougelsum: 0.3929
• Gen Len: 89.8971

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
2.2909	0.2786	10000	1.9040	0.4197	0.1915	0.2803	0.39	95.844
2.2169	0.5573	20000	1.8850	0.4166	0.1886	0.2831	0.3876	86.0059
2.1929	0.8359	30000	1.8784	0.4213	0.192	0.2858	0.391	86.0011
2.1806	1.1145	40000	1.8755	0.419	0.1905	0.2829	0.3892	91.4016
2.1683	1.3931	50000	1.8673	0.4202	0.1916	0.2834	0.3916	94.264
2.164	1.6718	60000	1.8603	0.4214	0.1923	0.2846	0.3919	90.9084
2.1594	1.9504	70000	1.8519	0.422	0.193	0.2864	0.3929	89.8971

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.6.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1