bart-base-aeslc-sentence-paraphrased-rouge-3-loss-differentiable-0-cnt-supervised-sequential

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: 7.8531
• Rouge1: 0.0861
• Rouge2: 0.0345
• Rougel: 0.0773
• Rougelsum: 0.0773

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
0.7003	0.0055	5	7.6959	0.0859	0.0342	0.0772	0.0771
0.6483	0.0111	10	7.7064	0.0859	0.0342	0.0772	0.0771
0.6934	0.0166	15	7.7242	0.0857	0.0342	0.077	0.077
0.6628	0.0222	20	7.7494	0.0855	0.034	0.0768	0.0767
0.6795	0.0277	25	7.7819	0.0855	0.034	0.0767	0.0766
0.6343	0.0333	30	7.8191	0.0856	0.0343	0.0768	0.0768
0.6625	0.0388	35	7.8531	0.0861	0.0345	0.0773	0.0773
0.6234	0.0444	40	7.8813	0.0856	0.034	0.0769	0.0769
0.6383	0.0499	45	7.8967	0.0851	0.0337	0.0765	0.0765
0.6694	0.0554	50	7.9148	0.085	0.0338	0.0765	0.0764
0.5896	0.0610	55	7.9225	0.085	0.0338	0.0765	0.0764
0.6354	0.0665	60	7.9385	0.0849	0.0339	0.0763	0.0763
0.6173	0.0721	65	7.9449	0.0846	0.0338	0.0761	0.0761
0.637	0.0776	70	7.9451	0.0846	0.0337	0.0758	0.0757
0.6285	0.0832	75	7.9347	0.0843	0.0336	0.0757	0.0757
0.6109	0.0887	80	7.9189	0.0841	0.0333	0.0754	0.0754
0.6405	0.0943	85	7.9218	0.0838	0.0334	0.0754	0.0753

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1