metadata

license: mit
tags:
  - generated_from_trainer
datasets:
  - billsum
metrics:
  - rouge
model-index:
  - name: bart-large-cnn-small-billsum-3epochs
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: billsum
          type: billsum
          config: default
          split: train
          args: default
        metrics:
          - name: Rouge1
            type: rouge
            value: 0.5409

bart-large-cnn-small-billsum-3epochs

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

Loss: 1.7523
Rouge1: 0.5409
Rouge2: 0.3112
Rougel: 0.3929
Rougelsum: 0.4633

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: 2.5764683748161164e-05
train_batch_size: 4
eval_batch_size: 8
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: 16
num_epochs: 3
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.7132	0.32	8	2.2000	0.4619	0.2328	0.3201	0.3939
2.236	0.64	16	1.9705	0.499	0.2768	0.3651	0.4216
2.1109	0.96	24	1.8845	0.5214	0.2974	0.3844	0.4417
1.7663	1.28	32	1.8211	0.5226	0.2935	0.3718	0.4479
1.7838	1.6	40	1.7981	0.5338	0.3001	0.383	0.4466
1.5229	1.92	48	1.7625	0.5299	0.3012	0.3839	0.4494
1.5221	2.24	56	1.7532	0.5384	0.3117	0.3939	0.4637
1.2879	2.56	64	1.7560	0.5338	0.3075	0.3865	0.4584
1.4046	2.88	72	1.7523	0.5409	0.3112	0.3929	0.4633

Framework versions

Transformers 4.25.1
Pytorch 1.13.0+cu117
Datasets 2.7.1
Tokenizers 0.13.2