t5-small-finetuned-t5-summarization

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

• Loss: 1.7613
• Rouge1: 24.5755
• Rouge2: 11.8424
• Rougel: 20.3031
• Rougelsum: 23.1867
• Gen Len: 18.9999

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: 6e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.9891	1.0	17945	1.7981	24.382	11.7099	20.1707	23.0021	18.9998
1.9527	2.0	35890	1.7816	24.4884	11.7673	20.2698	23.1233	19.0
1.9421	3.0	53835	1.7728	24.5782	11.8401	20.3343	23.2033	18.9997
1.9298	4.0	71780	1.7677	24.566	11.8723	20.3296	23.1943	18.9999
1.9256	5.0	89725	1.7619	24.5662	11.8385	20.3265	23.2016	18.9999
1.9056	6.0	107670	1.7613	24.5755	11.8424	20.3031	23.1867	18.9999

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2