mt5-small-finetuned-arxiv-cs

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

• Loss: 1.6922
• Rouge1: 0.7734
• Rouge2: 0.2865
• Rougel: 0.6665
• Rougelsum: 0.6743

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: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
14.0947	1.0	500	2.7666	1.2101	0.459	1.1426	1.1385
2.8524	2.0	1000	1.8208	0.0	0.0	0.0	0.0
2.2623	3.0	1500	1.6922	0.7734	0.2865	0.6665	0.6743

Framework versions

• Transformers 4.12.5
• Pytorch 1.10.0+cu111
• Datasets 1.15.1
• Tokenizers 0.10.3