marcelomoreno26/bart-large-samsum-adapter

bart-large-samsum-2

This model is a fine-tuned version of facebook/bart-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4648
• Rouge1: 0.4729
• Rouge2: 0.2361
• Rougel: 0.3953
• Rougelsum: 0.3947
• Gen Len: 18.0

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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	Gen Len
No log	1.0	460	1.5889	0.4523	0.2142	0.3714	0.3708	18.0
2.2048	2.0	921	1.5293	0.4642	0.231	0.3875	0.3871	18.0
1.765	3.0	1381	1.4971	0.4662	0.2268	0.3864	0.3857	18.0
1.7019	4.0	1842	1.4893	0.471	0.2337	0.3934	0.3925	18.0
1.6734	5.0	2302	1.4844	0.4725	0.2338	0.3945	0.3937	18.0
1.6536	6.0	2763	1.4707	0.4717	0.2341	0.3935	0.3928	18.0
1.6493	7.0	3223	1.4746	0.4736	0.2357	0.3956	0.3947	18.0
1.6363	8.0	3684	1.4688	0.4731	0.2344	0.3937	0.393	18.0
1.6337	9.0	4144	1.4658	0.4725	0.2345	0.3937	0.393	18.0
1.6283	9.99	4600	1.4648	0.4729	0.2361	0.3953	0.3947	18.0

Framework versions

• PEFT 0.10.1.dev0
• Transformers 4.39.3
• Pytorch 2.2.2
• Datasets 2.18.0
• Tokenizers 0.15.2