deberta-v3-large-survey-main_passage_old_facts-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: 0.9958
• Krippendorff: 0.8010
• Spearman: 0.8510
• Absolute Agreement: 0.6286
• Agreement Within One: 0.875

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	52	1.9583	0.0604	0.1334	0.0556	0.6111
No log	2.0	104	1.9600	0.0479	0.2283	0.0556	0.6111
No log	3.0	156	1.9760	-0.3323	-0.2987	0.0278	0.75
No log	4.0	208	2.1437	-0.5495	nan	0.1111	1.0
No log	5.0	260	2.1877	-0.5495	nan	0.1111	1.0
No log	6.0	312	2.2121	-0.5495	nan	0.1111	1.0
No log	7.0	364	2.2135	-0.5495	nan	0.1111	1.0
No log	8.0	416	2.1887	-0.4448	-0.0544	0.1111	0.9722
No log	9.0	468	2.0968	-0.0169	0.0989	0.1667	0.8333
1.683	10.0	520	2.1252	0.0707	0.1072	0.1389	0.7222
1.683	11.0	572	2.0715	0.2038	0.1796	0.1944	0.6944
1.683	12.0	624	2.2103	0.2636	0.2444	0.1944	0.7222
1.683	13.0	676	2.2643	0.2004	0.1892	0.1944	0.7222
1.683	14.0	728	2.4782	0.1607	0.0971	0.1389	0.75
1.683	15.0	780	2.4848	0.0716	0.0545	0.1111	0.6389
1.683	16.0	832	2.5613	0.1815	0.1711	0.1389	0.6944
1.683	17.0	884	2.6313	0.2082	0.2297	0.1667	0.7222
1.683	18.0	936	2.7220	0.1543	0.1385	0.1389	0.75
1.683	19.0	988	2.8428	0.2100	0.1983	0.1389	0.75
0.9272	20.0	1040	2.9826	0.2266	0.2310	0.1389	0.75

Framework versions

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