bart-paraphrase-finetuned-xsum-v3

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

• Loss: 0.3377
• Rouge1: 99.9461
• Rouge2: 72.6619
• Rougel: 99.9461
• Rougelsum: 99.9461
• Gen Len: 9.0396

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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	139	0.3653	96.4972	70.8271	96.5252	96.5085	9.7158
No log	2.0	278	0.6624	98.3228	72.2829	98.2598	98.2519	9.0612
No log	3.0	417	0.2880	98.2415	72.36	98.249	98.2271	9.4496
0.5019	4.0	556	0.4188	98.1123	70.8536	98.0746	98.0465	9.4065
0.5019	5.0	695	0.3718	98.8882	72.6619	98.8997	98.8882	10.7842
0.5019	6.0	834	0.4442	99.6076	72.6619	99.6076	99.598	9.0647
0.5019	7.0	973	0.2681	99.6076	72.6619	99.598	99.598	9.1403
0.2751	8.0	1112	0.3577	99.2479	72.6619	99.2536	99.2383	9.0612
0.2751	9.0	1251	0.2481	98.8785	72.6394	98.8882	98.8882	9.7914
0.2751	10.0	1390	0.2339	99.6076	72.6619	99.6076	99.6076	9.1942
0.2051	11.0	1529	0.2472	99.9461	72.6619	99.9461	99.9461	9.2338
0.2051	12.0	1668	0.3948	99.6076	72.6619	99.598	99.598	9.0468
0.2051	13.0	1807	0.4756	99.6076	72.6619	99.6076	99.6076	9.0576
0.2051	14.0	1946	0.3543	99.9461	72.6619	99.9461	99.9461	9.0396
0.1544	15.0	2085	0.2828	99.9461	72.6619	99.9461	99.9461	9.0576
0.1544	16.0	2224	0.2456	99.9461	72.6619	99.9461	99.9461	9.1079
0.1544	17.0	2363	0.2227	99.9461	72.6394	99.9461	99.9461	9.5072
0.1285	18.0	2502	0.3490	99.9461	72.6619	99.9461	99.9461	9.0396
0.1285	19.0	2641	0.3736	99.9461	72.6619	99.9461	99.9461	9.0396
0.1285	20.0	2780	0.3377	99.9461	72.6619	99.9461	99.9461	9.0396

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1