cnn_dailymail-summarization-t5-small-2022-09-08

This model is a fine-tuned version of t5-small on the cnn_dailymail 3.0.0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.6455
• Rouge1: 41.4235
• Rouge2: 19.0263
• Rougel: 29.2892
• Rougelsum: 38.6338
• Gen Len: 73.7273

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: 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: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.8435	0.28	10000	1.6998	24.3321	11.599	20.1028	22.9562	18.9997
1.8464	0.56	20000	1.6814	24.4483	11.6789	20.1798	23.0508	18.9996
1.8332	0.84	30000	1.6738	24.5531	11.7949	20.2834	23.1588	18.9994
1.8054	1.11	40000	1.6636	24.6194	11.843	20.3375	23.2259	18.9991
1.7958	1.39	50000	1.6597	24.5017	11.7755	20.2439	23.1148	18.9998
1.8095	1.67	60000	1.6546	24.5126	11.8043	20.2603	23.1175	18.9999
1.8127	1.95	70000	1.6521	24.4845	11.8136	20.2557	23.1089	18.9999
1.7952	2.23	80000	1.6488	24.6217	11.8877	20.3555	23.2514	18.9996
1.7863	2.51	90000	1.6477	24.5616	11.8489	20.3021	23.1754	18.9996
1.7824	2.79	100000	1.6464	24.5852	11.8531	20.3172	23.2089	18.9998

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.12.1+cu102
• Datasets 2.4.0
• Tokenizers 0.12.1