deberta-v3-large-survey-main_passage_old_facts-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: 0.3911
• Krippendorff: 0.9259
• Spearman: 0.9416
• Absolute Agreement: 0.8802
• Agreement Within One: 0.9343

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	1.9135	0.0331	0.0620	0.1944	0.7639
No log	2.0	110	1.9089	-0.2254	nan	0.2917	0.5556
No log	3.0	165	1.9020	-0.2254	nan	0.2917	0.5556
No log	4.0	220	1.9002	-0.1927	-0.1379	0.2639	0.5833
No log	5.0	275	2.0013	-0.2851	-0.0147	0.1667	0.9028
No log	6.0	330	1.9979	-0.2985	-0.0683	0.1667	0.875
No log	7.0	385	2.0259	-0.1959	0.0825	0.2083	0.9306
No log	8.0	440	1.8934	0.2597	0.2421	0.2361	0.6806
No log	9.0	495	1.8360	0.2956	0.3313	0.25	0.7222
1.7262	10.0	550	1.7921	0.3751	0.4065	0.3333	0.7639
1.7262	11.0	605	1.7312	0.4620	0.4217	0.3472	0.75
1.7262	12.0	660	1.7035	0.4364	0.4636	0.3194	0.7083
1.7262	13.0	715	1.7118	0.3978	0.4790	0.375	0.6528
1.7262	14.0	770	1.6110	0.5927	0.6020	0.4306	0.7778
1.7262	15.0	825	1.6626	0.4402	0.4920	0.4028	0.6944
1.7262	16.0	880	1.6148	0.5882	0.6199	0.375	0.7917
1.7262	17.0	935	1.6236	0.5597	0.6186	0.4167	0.75
1.7262	18.0	990	1.6352	0.4110	0.4575	0.3611	0.7222
1.0082	19.0	1045	1.5764	0.4562	0.5757	0.4028	0.6944
1.0082	20.0	1100	1.5087	0.5681	0.6098	0.4444	0.7639

Framework versions

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