bart-samsum

This model is a fine-tuned version of facebook/bart-large-xsum on the samsum dataset. It achieves the following results on the evaluation set:

• Loss: 1.3852
• Rouge1: 0.547
• Rouge2: 0.2837
• Rougel: 0.4462
• Rougelsum: 0.4454
• Gen Len: 29.72

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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.5201	0.27	500	1.4589	0.5276	0.2694	0.4246	0.424	33.5067
1.3757	0.54	1000	1.5105	0.506	0.2566	0.415	0.4146	29.76
1.3496	0.81	1500	1.4039	0.5365	0.2759	0.4233	0.4221	29.8
1.094	1.09	2000	1.4119	0.5407	0.2827	0.4293	0.4288	29.84
1.1488	1.36	2500	1.3680	0.5275	0.2637	0.423	0.4224	26.92
1.1222	1.63	3000	1.2875	0.5369	0.2844	0.4473	0.4463	29.2267
1.1092	1.9	3500	1.3968	0.533	0.2818	0.4354	0.4363	30.0667
0.8509	2.17	4000	1.3682	0.5306	0.2874	0.4327	0.4331	29.1467
0.9565	2.44	4500	1.3450	0.5466	0.2782	0.4419	0.4409	29.2133
0.8496	2.72	5000	1.3768	0.5366	0.2807	0.4359	0.4351	30.7733
0.8397	2.99	5500	1.3852	0.547	0.2837	0.4462	0.4454	29.72

Framework versions

• Transformers 4.33.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3