README.md

metadata

tags:
  - generated_from_trainer
datasets:
  - null
metrics:
  - rouge
model_index:
  - name: bart-base-finetuned-xsum
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        metric:
          name: Rouge1
          type: rouge
          value: 27.8379

bart-base-finetuned-xsum

This model is a fine-tuned version of facebook/bart-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5671
• Rouge1: 27.8379
• Rouge2: 16.2683
• Rougel: 24.1898
• Rougelsum: 25.5234
• Gen Len: 20.0

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.9931	1.0	879	1.5671	27.8379	16.2683	24.1898	25.5234	20.0

Framework versions

• Transformers 4.9.2
• Pytorch 1.9.0+cu102
• Datasets 1.11.0
• Tokenizers 0.10.3