deberta-v3-large-survey-new_fact_main_passage-rater

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: 0.2742
• Krippendorff: 0.9302
• Spearman: 0.9541
• Absolute Agreement: 0.9183
• Agreement Within One: 0.9837

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	2.0683	-0.3510	nan	0.0972	1.0
No log	2.0	100	2.0617	-0.3510	nan	0.0972	1.0
No log	3.0	150	2.0480	-0.3510	nan	0.0972	1.0
No log	4.0	200	1.9377	-0.5105	nan	0.2222	1.0
No log	5.0	250	2.0281	-0.5105	nan	0.2222	1.0
No log	6.0	300	2.1102	-0.5105	nan	0.2222	1.0
No log	7.0	350	2.1711	-0.1354	-0.0833	0.2361	0.7639
No log	8.0	400	2.2375	0.0597	0.1749	0.2917	0.9167
No log	9.0	450	2.2094	0.2618	0.2157	0.3194	0.8611
1.4101	10.0	500	2.2945	0.3359	0.3103	0.3611	0.8611
1.4101	11.0	550	2.0979	0.4477	0.3534	0.3611	0.8333
1.4101	12.0	600	2.0983	0.5901	0.5467	0.3472	0.8333
1.4101	13.0	650	2.4303	0.3729	0.2982	0.375	0.9167
1.4101	14.0	700	2.2451	0.7466	0.6756	0.4722	0.8611
1.4101	15.0	750	2.5756	0.5188	0.4433	0.4167	0.8889
1.4101	16.0	800	2.9836	0.4981	0.3828	0.3889	0.8889
1.4101	17.0	850	2.4424	0.8183	0.7631	0.4444	0.8889
1.4101	18.0	900	2.8010	0.7666	0.7242	0.4861	0.8611
1.4101	19.0	950	3.0376	0.6291	0.5873	0.4861	0.8333
0.4737	20.0	1000	3.3937	0.5765	0.5255	0.4306	0.9167

Framework versions

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