bart-abs-1509-0313-lr-3e-06-bs-8-maxep-10

This model is a fine-tuned version of sshleifer/distilbart-xsum-12-6 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 6.7933
• Rouge/rouge1: 0.2524
• Rouge/rouge2: 0.0705
• Rouge/rougel: 0.2098
• Rouge/rougelsum: 0.2104
• Bertscore/bertscore-precision: 0.8386
• Bertscore/bertscore-recall: 0.8631
• Bertscore/bertscore-f1: 0.8506
• Meteor: 0.2495
• Gen Len: 47.8636

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: 3e-06
• 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge/rouge1	Rouge/rouge2	Rouge/rougel	Rouge/rougelsum	Bertscore/bertscore-precision	Bertscore/bertscore-recall	Bertscore/bertscore-f1	Meteor	Gen Len
0.2585	1.0	109	6.7188	0.323	0.081	0.2516	0.2521	0.8764	0.8641	0.8702	0.2583	36.0
0.3824	2.0	218	6.7602	0.2515	0.0697	0.2094	0.21	0.8374	0.8631	0.85	0.2474	48.0
0.3739	3.0	327	6.7409	0.3111	0.0793	0.2212	0.2213	0.8659	0.864	0.8649	0.228	36.0
0.3565	4.0	436	6.7314	0.2439	0.0504	0.2065	0.2067	0.8544	0.8581	0.8562	0.229	45.0
0.3514	5.0	545	6.7648	0.2553	0.0701	0.2111	0.2116	0.839	0.8634	0.8509	0.2466	47.4545
0.3434	6.0	654	6.7689	0.271	0.0725	0.2029	0.2034	0.8612	0.8618	0.8614	0.2584	45.0
0.335	7.0	763	6.7845	0.2722	0.0714	0.2029	0.2031	0.8612	0.8618	0.8615	0.2582	45.0
0.337	8.0	872	6.7870	0.2601	0.0623	0.1995	0.1999	0.8579	0.8595	0.8586	0.2431	45.0
0.3328	9.0	981	6.7950	0.2515	0.0697	0.2094	0.21	0.8374	0.8631	0.85	0.2474	48.0
0.34	10.0	1090	6.7933	0.2524	0.0705	0.2098	0.2104	0.8386	0.8631	0.8506	0.2495	47.8636

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1