mt5-small-multi-news

This model is a fine-tuned version of google/mt5-small on the multi_news dataset. It achieves the following results on the evaluation set:

• Loss: 3.2170
• Rouge1: 22.03
• Rouge2: 6.95
• Rougel: 18.41
• Rougelsum: 18.72

Intended uses & limitations

Text summarization is the inteded use of this model. With further training the model could achieve better results.

Training and evaluation data

For the training data we used 10000 samples from the multi-news train dataset. For the evaluation data we used 500 samples from the multi-news evaluation dataset.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5.6e-05
• train_batch_size: 8
• eval_batch_size: 8
• 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
5.2732	1.0	1250	3.2170	22.03	6.95	18.41	18.72

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.0+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3