Modelpublic

domenicrosati/deberta-v3-large-survey-cross_passage_consistency-rater-all-gpt4

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes12downloads

Model Card

deberta-v3-large-survey-crosspassageconsistency-rater-all-gpt4

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: 0.4020
—Krippendorff: 0.8066
—Spearman: 0.8930
—Absolute Agreement: 0.8848
—Agreement Within One: 0.9447

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
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 20
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Krippendorff	Spearman	Absolute Agreement	Agreement Within One
No log	1.0	55	1.8997	-0.2987	0.2073	0.0556	0.9028
No log	2.0	110	1.8795	-0.3017	-0.0377	0.0833	0.9028
No log	3.0	165	1.8738	-0.0086	0.0516	0.2778	0.7778
No log	4.0	220	1.8663	-0.0943	-0.0377	0.3472	0.7361
No log	5.0	275	1.8618	-0.0943	-0.0377	0.3472	0.7361
No log	6.0	330	2.0154	-0.2704	nan	0.3194	0.6667
No log	7.0	385	1.9047	-0.2704	nan	0.3194	0.6667
No log	8.0	440	1.7587	-0.2704	nan	0.3194	0.6667
No log	9.0	495	1.6830	-0.2704	nan	0.3194	0.6667
1.409	10.0	550	1.6436	-0.2704	nan	0.3194	0.6667
1.409	11.0	605	1.6273	-0.2704	nan	0.3194	0.6667
1.409	12.0	660	1.5649	-0.2505	0.1100	0.3333	0.6667
1.409	13.0	715	1.5027	-0.0705	0.4448	0.4167	0.6806
1.409	14.0	770	1.5082	0.0375	0.4258	0.4028	0.6944
1.409	15.0	825	1.3945	0.0492	0.5924	0.4861	0.7083
1.409	16.0	880	1.3575	0.0458	0.5905	0.4861	0.6806
1.409	17.0	935	1.3444	0.0492	0.5924	0.4861	0.7083
1.409	18.0	990	1.2574	0.1791	0.5461	0.5278	0.7778
0.7706	19.0	1045	1.2676	0.0357	0.5486	0.5139	0.7361
0.7706	20.0	1100	1.2345	0.0376	0.5548	0.5278	0.7222

Framework versions

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