deberta-v3-large-survey-related_passage_old_facts-rater-all-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: 0.4609
• Krippendorff: 0.9507
• Spearman: 0.9543
• Absolute Agreement: 0.8790
• Agreement Within One: 0.9424

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	1.9175	-0.5892	-0.2660	0.0833	0.9583
No log	2.0	110	1.9155	-0.6559	nan	0.0972	1.0
No log	3.0	165	1.9163	-0.6559	nan	0.0972	1.0
No log	4.0	220	1.9146	-0.6559	nan	0.0972	1.0
No log	5.0	275	1.9040	-0.2400	-0.1609	0.1389	0.7778
No log	6.0	330	1.8582	-0.3203	nan	0.2778	0.5556
No log	7.0	385	1.8332	-0.3203	nan	0.2778	0.5556
No log	8.0	440	1.7830	0.4020	0.4309	0.3472	0.7222
No log	9.0	495	1.7201	0.4242	0.4465	0.3472	0.7083
1.7718	10.0	550	1.7090	0.4242	0.4465	0.3472	0.7083
1.7718	11.0	605	1.7395	0.4927	0.4940	0.3611	0.7361
1.7718	12.0	660	1.6263	0.4151	0.4645	0.3472	0.6806
1.7718	13.0	715	1.5841	0.4799	0.5312	0.4028	0.7222
1.7718	14.0	770	1.5878	0.4969	0.4540	0.3611	0.75
1.7718	15.0	825	1.5259	0.5079	0.5361	0.4028	0.7361
1.7718	16.0	880	1.7216	0.5194	0.5553	0.3889	0.75
1.7718	17.0	935	1.6601	0.4413	0.5244	0.4028	0.6944
1.7718	18.0	990	1.6167	0.5381	0.5698	0.4722	0.7639
1.0304	19.0	1045	1.8458	0.4673	0.5520	0.4306	0.7083
1.0304	20.0	1100	1.5816	0.5537	0.6478	0.5278	0.7778

Framework versions

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