mt5-small-finetune-sumsum

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

• Loss: 3.3263
• Rouge1: 20.9651
• Rouge2: 7.1527
• Rougel: 18.4396
• Rougelsum: 19.5209

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: 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: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
12.2648	1.0	125	4.3790	9.8078	1.7255	9.0852	9.4233
6.0853	2.0	250	3.4753	20.7185	6.502	18.079	19.2584
4.9838	3.0	375	3.3263	20.9651	7.1527	18.4396	19.5209

Framework versions

• Transformers 4.18.0
• Pytorch 1.13.0
• Datasets 2.6.1
• Tokenizers 0.11.0