deberta-v3-large-survey-new_fact_related_passage-rater-all

This model is a fine-tuned version of microsoft/deberta-v3-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4933
• Krippendorff: 0.8215
• Spearman: 0.8423
• Absolute Agreement: 0.8468
• Agreement Within One: 0.9389

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: 6e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Krippendorff	Spearman	Absolute Agreement	Agreement Within One
No log	1.0	55	1.9220	-0.5318	nan	0.1806	1.0
No log	2.0	110	1.9208	-0.5318	nan	0.1806	1.0
No log	3.0	165	1.9200	-0.5318	nan	0.1806	1.0
No log	4.0	220	1.9197	-0.5318	nan	0.1806	1.0
No log	5.0	275	1.9655	-0.5318	nan	0.1806	1.0
No log	6.0	330	1.9944	-0.4584	-0.1731	0.1528	0.9167
No log	7.0	385	1.9699	-0.3160	-0.3151	0.1389	0.6944
No log	8.0	440	1.8888	-0.3710	-0.3364	0.1667	0.6806
No log	9.0	495	1.8273	0.1437	0.2775	0.375	0.625
1.6819	10.0	550	1.7380	-0.0391	0.1674	0.3472	0.5833
1.6819	11.0	605	1.6954	-0.3105	0.0146	0.3333	0.5417
1.6819	12.0	660	1.6211	0.3129	0.3446	0.4028	0.6806
1.6819	13.0	715	1.5305	0.2303	0.4155	0.4028	0.625
1.6819	14.0	770	1.4929	0.3973	0.4054	0.4306	0.7083
1.6819	15.0	825	1.4731	0.3220	0.3517	0.4167	0.6806
1.6819	16.0	880	1.4963	0.0524	0.3087	0.3889	0.5972
1.6819	17.0	935	1.4038	0.3960	0.4802	0.4722	0.6944
1.6819	18.0	990	1.3351	0.5195	0.4801	0.5139	0.7778
0.9585	19.0	1045	1.3013	0.4739	0.5167	0.5139	0.7361
0.9585	20.0	1100	1.2788	0.4785	0.4990	0.4861	0.7361

Framework versions

• Transformers 4.26.0
• Pytorch 1.13.1
• Datasets 2.10.1
• Tokenizers 0.12.1