Modelpublic

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

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes14downloads

Model Card

deberta-v3-large-survey-mainpassageconsistency-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: 0.3709
—Krippendorff: 0.7891
—Spearman: 0.8055
—Absolute Agreement: 0.9002
—Agreement Within One: 0.9519

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	1.9290	-0.7901	nan	0.0556	1.0
No log	2.0	104	1.9260	-0.7901	nan	0.0556	1.0
No log	3.0	156	2.2136	-0.3366	nan	0.1667	0.8056
No log	4.0	208	3.0444	-0.3366	nan	0.1667	0.8056
No log	5.0	260	3.3406	-0.3366	nan	0.1667	0.8056
No log	6.0	312	3.6074	-0.3366	nan	0.1667	0.8056
No log	7.0	364	3.6395	-0.3366	nan	0.1667	0.8056
No log	8.0	416	3.5991	0.2720	0.1638	0.1389	0.8611
No log	9.0	468	3.7407	0.2147	0.1298	0.1389	0.8611
0.9989	10.0	520	3.8742	0.2347	0.2208	0.1389	0.8889
0.9989	11.0	572	3.7369	0.3144	0.3601	0.1667	0.8889
0.9989	12.0	624	3.9522	0.2347	0.2208	0.1389	0.8889
0.9989	13.0	676	3.4874	0.2598	0.3174	0.1944	0.8889
0.9989	14.0	728	4.1072	0.2420	0.1444	0.1389	0.8611
0.9989	15.0	780	4.6448	0.3090	0.3144	0.1389	0.9167
0.9989	16.0	832	4.2375	0.3130	0.2379	0.1389	0.9167
0.9989	17.0	884	4.4929	0.1724	0.1068	0.1111	0.9167
0.9989	18.0	936	4.2319	0.2867	0.1237	0.1389	0.8889
0.9989	19.0	988	4.3529	0.2572	0.0514	0.1111	0.8889
0.3366	20.0	1040	4.5277	0.2274	0.0428	0.1111	0.8611

Framework versions

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