deberta-v3-large-survey-related_passage_old_facts-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: 1.0656
• Krippendorff: 0.7684
• Spearman: 0.8108
• Absolute Agreement: 0.6168
• Agreement Within One: 0.8492

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	1.9401	-0.6559	nan	0.0972	1.0
No log	2.0	100	1.9396	-0.6559	nan	0.0972	1.0
No log	3.0	150	1.9441	-0.6559	nan	0.0972	1.0
No log	4.0	200	1.9696	-0.6559	nan	0.0972	1.0
No log	5.0	250	1.9797	-0.6559	nan	0.0972	1.0
No log	6.0	300	1.9885	-0.4936	-0.0607	0.0972	0.9583
No log	7.0	350	2.0154	-0.3921	-0.2733	0.1111	0.8472
No log	8.0	400	2.0354	-0.1854	-0.1735	0.1389	0.8056
No log	9.0	450	2.1847	-0.4444	-0.1842	0.125	0.9167
1.6791	10.0	500	2.1343	-0.1919	-0.0366	0.1667	0.875
1.6791	11.0	550	2.3802	-0.2199	0.0219	0.1667	0.9028
1.6791	12.0	600	2.4238	-0.1432	0.0627	0.1944	0.8889
1.6791	13.0	650	2.1645	0.1016	0.2078	0.2778	0.8194
1.6791	14.0	700	2.3498	0.0959	0.2604	0.2639	0.875
1.6791	15.0	750	2.4257	0.0430	0.2141	0.2083	0.8889
1.6791	16.0	800	2.6584	0.0886	0.2120	0.1944	0.8889
1.6791	17.0	850	2.8723	-0.0019	0.1364	0.1806	0.875
1.6791	18.0	900	2.7402	0.0670	0.1728	0.2222	0.8194
1.6791	19.0	950	3.1369	-0.0312	0.1278	0.1528	0.8611
0.8687	20.0	1000	3.2599	-0.0863	0.0656	0.2083	0.8333

Framework versions

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