my_awesome_billsum_model

This model is a fine-tuned version of t5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0467
• Rouge1: 0.7832
• Rouge2: 0.692
• Rougel: 0.781
• Rougelsum: 0.7805
• Gen Len: 11.6071

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: 2e-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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	56	4.9294	0.0	0.0	0.0	0.0	16.558
No log	2.0	112	2.2288	0.0	0.0	0.0	0.0	13.5357
No log	3.0	168	0.4763	0.0045	0.0045	0.0045	0.0045	10.6518
No log	4.0	224	0.1138	0.7232	0.6205	0.7245	0.7236	11.5893
No log	5.0	280	0.0654	0.7417	0.6339	0.7417	0.7402	11.6607
No log	6.0	336	0.0587	0.7321	0.6205	0.7321	0.7309	11.5938
No log	7.0	392	0.0552	0.7496	0.6473	0.7491	0.7491	11.625
No log	8.0	448	0.0533	0.7714	0.6786	0.7714	0.7709	11.6562
1.6431	9.0	504	0.0518	0.781	0.692	0.7832	0.7805	11.6161
1.6431	10.0	560	0.0505	0.764	0.6652	0.7632	0.7614	11.6607
1.6431	11.0	616	0.0494	0.7778	0.6875	0.78	0.7773	11.6116
1.6431	12.0	672	0.0488	0.7778	0.6875	0.78	0.7773	11.6116
1.6431	13.0	728	0.0483	0.781	0.692	0.7815	0.7805	11.6161
1.6431	14.0	784	0.0479	0.781	0.692	0.7815	0.7805	11.6071
1.6431	15.0	840	0.0475	0.7852	0.6964	0.7839	0.7842	11.6205
1.6431	16.0	896	0.0471	0.7812	0.692	0.781	0.7805	11.5982
1.6431	17.0	952	0.0469	0.7884	0.7009	0.7879	0.7869	11.625
0.062	18.0	1008	0.0468	0.7832	0.692	0.781	0.7805	11.6071
0.062	19.0	1064	0.0467	0.7864	0.6964	0.7839	0.7837	11.6027
0.062	20.0	1120	0.0467	0.7832	0.692	0.781	0.7805	11.6071

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0