deberta-v3-large-survey-new_fact_main_passage-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: 0.7963
• Krippendorff: 0.7388
• Spearman: 0.7347
• Absolute Agreement: 0.7961
• Agreement Within One: 0.9401

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	1.8840	-0.1735	0.1672	0.3194	0.5694
No log	2.0	110	1.8714	0.3325	0.2924	0.3472	0.8333
No log	3.0	165	1.8455	0.3427	0.4616	0.375	0.9722
No log	4.0	220	1.8345	-0.3980	0.1275	0.2361	1.0
No log	5.0	275	1.8686	0.5029	0.5400	0.4306	0.9306
No log	6.0	330	1.6423	0.8650	0.8167	0.5278	0.8889
No log	7.0	385	1.6656	0.8650	0.8167	0.5278	0.8889
No log	8.0	440	1.6436	0.8626	0.8162	0.5278	0.875
No log	9.0	495	1.5656	0.8626	0.8162	0.5278	0.875
1.2212	10.0	550	1.5328	0.7676	0.7513	0.5139	0.8611
1.2212	11.0	605	1.5906	0.8626	0.8162	0.5278	0.875
1.2212	12.0	660	1.3500	0.7795	0.7681	0.5278	0.8611
1.2212	13.0	715	1.1828	0.8324	0.7641	0.5972	0.875
1.2212	14.0	770	1.3495	0.7277	0.7280	0.5694	0.9028
1.2212	15.0	825	1.2211	0.8510	0.7811	0.6111	0.8889
1.2212	16.0	880	1.1597	0.8487	0.7714	0.625	0.875
1.2212	17.0	935	1.3074	0.8454	0.8281	0.625	0.875
1.2212	18.0	990	1.1712	0.8548	0.8057	0.625	0.875
0.4567	19.0	1045	1.0816	0.8484	0.8103	0.6389	0.875
0.4567	20.0	1100	1.0759	0.8537	0.8038	0.6111	0.8889

Framework versions

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