Edit model card

flan-t5-base-samsum

This model is a fine-tuned version of google/flan-t5-base on the samsum dataset. It achieves the following results on the evaluation set:

  • Loss: 1.3794
  • Rouge1: 46.8948
  • Rouge2: 23.4445
  • Rougel: 39.5763
  • Rougelsum: 43.209
  • Gen Len: 17.2540

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: 5e-05
  • train_batch_size: 64
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Gen Len
No log 1.0 231 1.3935 46.6142 23.0937 39.1018 42.8696 17.2076
No log 2.0 462 1.3848 46.5553 23.0122 39.1493 42.764 17.1465
1.4249 3.0 693 1.3813 46.8705 23.5239 39.6689 43.2545 17.2930
1.4249 4.0 924 1.3801 46.9726 23.6143 39.6028 43.3278 17.2112
1.3528 5.0 1155 1.3794 46.8948 23.4445 39.5763 43.209 17.2540

Framework versions

  • Transformers 4.27.3
  • Pytorch 1.10.0
  • Datasets 2.10.1
  • Tokenizers 0.13.2
Downloads last month
3
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train hongdoubao/flan-t5-base-samsum

Evaluation results