README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - eli5
metrics:
  - rouge
model-index:
  - name: t5-small-finetuned-eli5-neel
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: eli5
          type: eli5
          config: LFQA_reddit
          split: train_eli5
          args: LFQA_reddit
        metrics:
          - name: Rouge1
            type: rouge
            value: 9.613

t5-small-finetuned-eli5-neel

This model is a fine-tuned version of t5-small on the eli5 dataset. It achieves the following results on the evaluation set:

• Loss: 3.6887
• Rouge1: 9.613
• Rouge2: 1.7491
• Rougel: 8.8341
• Rougelsum: 9.3402
• Gen Len: 19.0

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: 16
• eval_batch_size: 16
• 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
3.896	1.0	17040	3.6887	9.613	1.7491	8.8341	9.3402	19.0

Framework versions

• Transformers 4.22.1
• Pytorch 1.12.1+cu113
• Datasets 2.5.1
• Tokenizers 0.12.1