BARTkrame-abstract

This model is a fine-tuned version of krm/BARTkrame-abstract on the krm/for-ULPGL-Dissertation dataset. It achieves (15/10/2022) the following results on the evaluation set:

• Loss: 2.4196
• Rouge1: 0.2703
• Rouge2: 0.1334
• Rougel: 0.2392
• Rougelsum: 0.2419

Model description

This model is primarly a finetuned version of moussaKam/mbarthez.

Intended uses & limitations

More information needed

Training and evaluation data

We have used the krm/for-ULPGL-Dissertation dataset reduced to :

Training data : 5000 samples taken at random with seed=42.

Validation data : 100 samples taken at random with seed=42.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5.6e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9, 0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 12

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
0.1316	9.0	1250	2.3251	0.2505	0.1158	0.2150	0.2184
0.0894	10.0	2500	2.3467	0.2526	0.1073	0.2067	0.2124
0.045	11.0	3750	2.3742	0.2593	0.1211	0.2281	0.2308
0.0242	12.0	5000	2.4196	0.2703	0.1334	0.2392	0.2419

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.1+cu113
• Datasets 2.6.1
• Tokenizers 0.13.1