2023_12_20_06_21_29

This model is a fine-tuned version of google/flan-t5-large on the background_summ dataset. It achieves the following results on the evaluation set:

• Loss: 2.9279
• Rouge1: 36.6
• Rouge2: 13.8
• Rougel: 22.9
• Rougelsum: 33.1
• Bertscore Precision: 85.7
• Bertscore Recall: 86.9
• Bertscore F1: 86.3

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Bertscore Precision	Bertscore Recall	Bertscore F1
No log	1.0	370	2.0078	31.9	12.6	21.2	28.8	83.6	85.5	84.5
1.72	2.0	740	2.2254	31.6	12.2	20.9	28.0	83.7	85.9	84.7
1.1992	3.0	1110	2.4543	34.1	12.9	22.1	30.5	84.8	86.3	85.5
1.1992	4.0	1480	2.5872	35.4	13.7	22.7	32.0	85.1	86.7	85.9
0.95	5.0	1850	2.7400	36.1	14.0	22.9	32.8	85.4	86.9	86.2
0.8276	6.0	2220	2.7785	36.7	14.0	23.0	33.2	85.7	86.9	86.3
0.7381	7.0	2590	2.8543	37.2	14.5	23.4	33.7	85.8	87.1	86.4
0.7381	8.0	2960	2.8794	37.1	14.2	23.3	33.5	85.9	87.0	86.4
0.7023	9.0	3330	2.8693	36.8	14.0	23.0	33.2	85.8	87.0	86.3
0.6779	10.0	3700	2.9279	36.6	13.8	22.9	33.1	85.7	86.9	86.3

Framework versions

• Transformers 4.33.1
• Pytorch 1.13.1
• Datasets 2.14.5
• Tokenizers 0.13.3