fine-tuned-bart-20-epochs-1024-input-192-output

This model is a fine-tuned version of bart-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1893
• Rouge1: 0.1939
• Rouge2: 0.0401
• Rougel: 0.1527
• Rougelsum: 0.1511
• Gen Len: 32.12

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: 0.0001
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	151	6.2262	0.0	0.0	0.0	0.0	8.25
No log	2.0	302	1.4729	0.0968	0.0293	0.0904	0.0904	12.22
No log	3.0	453	1.2456	0.0905	0.0238	0.0705	0.071	20.86
4.2193	4.0	604	1.1718	0.1322	0.0275	0.1021	0.1031	26.99
4.2193	5.0	755	1.1348	0.1503	0.0288	0.1184	0.1183	30.21
4.2193	6.0	906	1.1201	0.1448	0.0266	0.1198	0.1196	25.39
0.9427	7.0	1057	1.1005	0.1369	0.0366	0.1083	0.1087	32.5
0.9427	8.0	1208	1.0998	0.1568	0.0396	0.1159	0.1155	38.52
0.9427	9.0	1359	1.1039	0.1656	0.0292	0.137	0.1361	28.11
0.6448	10.0	1510	1.1099	0.1838	0.0359	0.1398	0.1393	33.68
0.6448	11.0	1661	1.1146	0.182	0.0369	0.1439	0.1419	36.05
0.6448	12.0	1812	1.1208	0.1861	0.0416	0.1455	0.1441	36.32
0.6448	13.0	1963	1.1359	0.1755	0.0314	0.1381	0.1365	33.6
0.444	14.0	2114	1.1549	0.1913	0.0411	0.1514	0.1508	36.25
0.444	15.0	2265	1.1661	0.1773	0.041	0.1381	0.1369	33.62
0.444	16.0	2416	1.1664	0.1787	0.0449	0.1422	0.1417	30.72
0.3149	17.0	2567	1.1806	0.1886	0.0409	0.1509	0.1499	32.3
0.3149	18.0	2718	1.1837	0.189	0.04	0.1544	0.1529	31.79
0.3149	19.0	2869	1.1897	0.1937	0.0398	0.1546	0.1534	32.04
0.2519	20.0	3020	1.1893	0.1939	0.0401	0.1527	0.1511	32.12

Framework versions

• Transformers 4.36.2
• Pytorch 1.12.1+cu113
• Datasets 2.16.1
• Tokenizers 0.15.1