deberta-v3-large-survey-new_fact_related_passage-rater-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: 0.2756
• Krippendorff: 0.9122
• Spearman: 0.9223
• Absolute Agreement: 0.9083
• Agreement Within One: 0.9774

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	50	1.9756	-0.1842	nan	0.0694	0.8194
No log	2.0	100	1.9780	-0.1842	nan	0.0694	0.8194
No log	3.0	150	1.9800	-0.1842	nan	0.0694	0.8194
No log	4.0	200	1.9820	-0.1842	nan	0.0694	0.8194
No log	5.0	250	2.1958	-0.0858	-0.1251	0.0278	0.7083
No log	6.0	300	2.2100	0.0339	0.1339	0.0556	0.7361
No log	7.0	350	2.2628	0.0091	0.2531	0.0972	0.8611
No log	8.0	400	2.2684	-0.2678	0.0357	0.1528	0.9167
No log	9.0	450	2.2697	-0.2678	0.0357	0.1528	0.9167
1.4007	10.0	500	2.5200	-0.1996	0.0693	0.125	0.875
1.4007	11.0	550	2.4621	-0.0241	0.1621	0.1667	0.8889
1.4007	12.0	600	2.7691	-0.0280	0.1054	0.1528	0.8611
1.4007	13.0	650	2.6624	0.0140	0.1642	0.2083	0.8889
1.4007	14.0	700	2.7726	0.0826	0.1383	0.1389	0.8056
1.4007	15.0	750	3.2655	0.2014	0.2089	0.1389	0.7778
1.4007	16.0	800	3.3104	0.0622	0.2035	0.1667	0.8611
1.4007	17.0	850	3.3013	0.2135	0.2567	0.1806	0.8333
1.4007	18.0	900	3.0050	0.2975	0.2830	0.1667	0.7778
1.4007	19.0	950	3.4558	0.1430	0.2280	0.1667	0.8611
0.5598	20.0	1000	3.1088	0.2953	0.2430	0.1944	0.7917

Framework versions

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