Modelpublic

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

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes9downloads

Model Card

deberta-v3-large-survey-crosspassageconsistency-rater-half-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: 0.5957
—Krippendorff: 0.5530
—Spearman: 0.6991
—Absolute Agreement: 0.8221
—Agreement Within One: 0.9002

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	52	1.9600	-0.4776	0.3301	0.0556	0.9722
No log	2.0	104	1.9258	-0.2795	0.2827	0.0833	0.9722
No log	3.0	156	1.9351	-0.3095	0.2873	0.0833	0.9722
No log	4.0	208	2.0079	-0.1471	-0.1057	0.0833	0.8333
No log	5.0	260	2.2469	-0.2860	nan	0.1944	0.8056
No log	6.0	312	2.2186	-0.2860	nan	0.1944	0.8056
No log	7.0	364	2.2266	-0.2860	nan	0.1944	0.8056
No log	8.0	416	2.2258	-0.2860	nan	0.1944	0.8056
No log	9.0	468	2.2048	-0.2860	nan	0.1944	0.8056
1.3796	10.0	520	2.2347	-0.2860	nan	0.1944	0.8056
1.3796	11.0	572	2.2480	-0.2860	nan	0.1944	0.8056
1.3796	12.0	624	2.1409	-0.2349	0.0	0.1944	0.8056
1.3796	13.0	676	2.0869	0.0452	0.1218	0.1944	0.8333
1.3796	14.0	728	2.1785	0.1597	0.2372	0.1944	0.8611
1.3796	15.0	780	2.1684	0.0452	0.1218	0.1944	0.8333
1.3796	16.0	832	2.3356	0.0185	0.1110	0.1944	0.8333
1.3796	17.0	884	2.5980	-0.0694	0.0783	0.2222	0.8333
1.3796	18.0	936	2.3700	0.0944	0.2085	0.1944	0.8611
1.3796	19.0	988	2.3355	-0.0396	0.0913	0.1667	0.8611
0.7039	20.0	1040	2.7088	-0.0625	0.0901	0.1667	0.8611

Framework versions

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