Chikashi's picture
update model card README.md
5c605c0
metadata
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - cnn_dailymail
metrics:
  - rouge
model-index:
  - name: t5-small-finetuned-cnndm3-wikihow2
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: cnn_dailymail
          type: cnn_dailymail
          args: 3.0.0
        metrics:
          - name: Rouge1
            type: rouge
            value: 24.6704

t5-small-finetuned-cnndm3-wikihow2

This model is a fine-tuned version of Chikashi/t5-small-finetuned-cnndm2-wikihow2 on the cnn_dailymail dataset. It achieves the following results on the evaluation set:

  • Loss: 1.6265
  • Rouge1: 24.6704
  • Rouge2: 11.9038
  • Rougel: 20.3622
  • Rougelsum: 23.2612
  • Gen Len: 18.9997

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: 0.0003
  • 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
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Gen Len
1.8071 1.0 71779 1.6265 24.6704 11.9038 20.3622 23.2612 18.9997

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.10.0+cu111
  • Datasets 2.1.0
  • Tokenizers 0.12.1