bart-cnn-dailymail-seed42

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

• Loss: 1.7126
• Rouge1: 0.2361
• Rouge2: 0.1189
• Rougel: 0.1981
• Rougelsum: 0.2238

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: 5e-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
• lr_scheduler_warmup_steps: 2000
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.2539	0.1115	1000	1.8901	0.2314	0.1104	0.1915	0.2177
2.1395	0.2229	2000	1.8539	0.2324	0.1110	0.1923	0.2186
2.0919	0.3344	3000	1.8197	0.2318	0.1137	0.1931	0.2186
2.0505	0.4458	4000	1.7912	0.2361	0.1178	0.1979	0.2234
2.0239	0.5573	5000	1.7712	0.2347	0.1174	0.1966	0.2221
1.9945	0.6687	6000	1.7501	0.2375	0.1187	0.1986	0.2246
1.9875	0.7802	7000	1.7361	0.2356	0.1183	0.1970	0.2225
1.9728	0.8916	8000	1.7404	0.2381	0.1193	0.1992	0.2252
1.9517	1.0031	9000	1.7336	0.2379	0.1191	0.1990	0.2249
1.8701	1.1145	10000	1.7226	0.2362	0.1186	0.1979	0.2233
1.8677	1.2260	11000	1.7126	0.2361	0.1189	0.1981	0.2238

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1