theojolliffe/bart-cnn-v3-e16

bart-large-cnn-finetuned-roundup-4-16

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

• Loss: 0.8760
• Rouge1: 56.3338
• Rouge2: 42.4032
• Rougel: 45.9455
• Rougelsum: 54.6488
• Gen Len: 142.0

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	398	0.9325	52.7796	33.0802	34.8217	50.2211	142.0
1.1317	2.0	796	0.8313	53.6274	35.3235	37.7077	51.0888	141.2963
0.6757	3.0	1194	0.7893	54.1449	34.7532	36.3211	51.781	142.0
0.4511	4.0	1592	0.7647	52.2694	34.2286	36.5736	49.7078	142.0
0.4511	5.0	1990	0.7596	55.1986	37.5865	41.406	53.1897	141.8333
0.3037	6.0	2388	0.7688	53.9367	36.8729	39.9456	51.5108	142.0
0.209	7.0	2786	0.7590	54.6867	37.6415	41.2602	52.746	142.0
0.1452	8.0	3184	0.7744	53.5374	36.3666	40.0432	51.3461	142.0
0.11	9.0	3582	0.8042	56.6623	40.4702	44.0028	54.5138	142.0
0.11	10.0	3980	0.8105	55.6002	40.5663	43.8119	53.9117	142.0
0.0833	11.0	4378	0.8230	56.2517	40.8567	44.0009	54.3271	142.0
0.0634	12.0	4776	0.8329	55.9228	40.6443	43.6161	54.0975	142.0
0.0474	13.0	5174	0.8570	55.4923	40.3683	43.4675	53.404	142.0
0.0349	14.0	5572	0.8658	56.4454	41.8069	44.2922	54.464	142.0
0.0349	15.0	5970	0.8754	56.3837	42.2025	45.7817	54.4912	142.0
0.0304	16.0	6368	0.8760	56.3338	42.4032	45.9455	54.6488	142.0

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 2.1.0
• Tokenizers 0.12.1