Modelpublic

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

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes8downloads

Model Card

deberta-v3-large-survey-newfactmain_passage-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: 0.6360
—Krippendorff: 0.8330
—Spearman: 0.8242
—Absolute Agreement: 0.8498
—Agreement Within One: 0.9435

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	1.9056	-0.1975	0.2732	0.1389	0.5556
No log	2.0	104	1.8986	-0.3052	0.0684	0.1389	0.5556
No log	3.0	156	1.8869	-0.3835	nan	0.1389	0.5278
No log	4.0	208	2.0502	-0.4830	nan	0.25	1.0
No log	5.0	260	2.4002	-0.4830	nan	0.25	1.0
No log	6.0	312	2.4895	-0.0731	0.0193	0.2222	0.8889
No log	7.0	364	2.2943	0.7464	0.7453	0.3889	0.8889
No log	8.0	416	2.2416	0.8118	0.7882	0.3889	0.8611
No log	9.0	468	2.3725	0.8118	0.7882	0.3889	0.8611
1.2529	10.0	520	2.4889	0.6015	0.6308	0.3333	0.7778
1.2529	11.0	572	2.7181	0.6015	0.6308	0.3333	0.7778
1.2529	12.0	624	2.5870	0.6829	0.6950	0.3611	0.8056
1.2529	13.0	676	2.7559	0.6400	0.6437	0.3611	0.8056
1.2529	14.0	728	2.8805	0.6829	0.6950	0.3611	0.8056
1.2529	15.0	780	2.7286	0.6814	0.6549	0.3889	0.8333
1.2529	16.0	832	3.0974	0.7322	0.7024	0.3889	0.8333
1.2529	17.0	884	3.2368	0.6794	0.6160	0.2778	0.8056
1.2529	18.0	936	3.1932	0.6414	0.5821	0.25	0.7778
1.2529	19.0	988	3.3644	0.7016	0.6197	0.3056	0.8333
0.4102	20.0	1040	3.6545	0.6146	0.5778	0.25	0.75

Framework versions

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