fine-tuned-BioBARTv2-20-epochs-1024-input-352-output

This model is a fine-tuned version of checkpoint_global_step_200000 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7137
• Rouge1: 0.1879
• Rouge2: 0.0399
• Rougel: 0.1477
• Rougelsum: 0.1473
• Gen Len: 37.71

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	6.2824	0.0873	0.0004	0.0696	0.0699	59.17
No log	2.0	302	0.8585	0.1021	0.0276	0.0905	0.0905	23.51
No log	3.0	453	0.7401	0.0672	0.0142	0.0529	0.0523	21.97
4.0066	4.0	604	0.6962	0.1224	0.0287	0.0968	0.0968	29.25
4.0066	5.0	755	0.6739	0.1497	0.0295	0.1199	0.1188	34.66
4.0066	6.0	906	0.6642	0.1548	0.0299	0.1156	0.1142	50.23
0.5957	7.0	1057	0.6592	0.1319	0.0281	0.0993	0.0979	37.47
0.5957	8.0	1208	0.6532	0.1756	0.0366	0.1416	0.1411	38.41
0.5957	9.0	1359	0.6604	0.1636	0.034	0.1298	0.1291	33.72
0.4198	10.0	1510	0.6624	0.1841	0.0389	0.1439	0.1423	37.58
0.4198	11.0	1661	0.6656	0.1864	0.0331	0.1479	0.1472	46.92
0.4198	12.0	1812	0.6683	0.1918	0.0426	0.1432	0.1432	45.94
0.4198	13.0	1963	0.6796	0.1851	0.0374	0.1396	0.1393	47.93
0.3012	14.0	2114	0.6847	0.1933	0.0393	0.1413	0.1407	41.22
0.3012	15.0	2265	0.6919	0.175	0.036	0.132	0.131	38.91
0.3012	16.0	2416	0.7011	0.1985	0.03	0.1495	0.1494	43.78
0.2208	17.0	2567	0.7098	0.1836	0.033	0.1395	0.1377	38.65
0.2208	18.0	2718	0.7080	0.1888	0.038	0.1433	0.1417	39.49
0.2208	19.0	2869	0.7127	0.186	0.0351	0.1479	0.1479	39.35
0.1823	20.0	3020	0.7137	0.1879	0.0399	0.1477	0.1473	37.71

Framework versions

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