deberta-v3-large-survey-new_fact_main_passage-rater-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: 0.5088
• Krippendorff: 0.8691
• Spearman: 0.8893
• Absolute Agreement: 0.8706
• Agreement Within One: 0.9108

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	1.9589	-0.0190	0.0341	0.2222	0.625
No log	2.0	100	1.9500	-0.1769	-0.0116	0.2639	0.5972
No log	3.0	150	1.8211	-0.5105	nan	0.2222	1.0
No log	4.0	200	2.0070	-0.5105	nan	0.2222	1.0
No log	5.0	250	2.0851	-0.5105	nan	0.2222	1.0
No log	6.0	300	2.2159	-0.5105	nan	0.2222	1.0
No log	7.0	350	1.8615	0.8096	0.7617	0.5139	0.8889
No log	8.0	400	1.8606	0.8096	0.7617	0.5139	0.8889
No log	9.0	450	1.8884	0.8650	0.8167	0.5278	0.8889
1.1332	10.0	500	1.9244	0.8650	0.8167	0.5278	0.8889
1.1332	11.0	550	2.0776	0.8650	0.8167	0.5278	0.8889
1.1332	12.0	600	2.0743	0.7895	0.7588	0.5139	0.9028
1.1332	13.0	650	2.4128	0.8650	0.8167	0.5278	0.8889
1.1332	14.0	700	2.3573	0.8650	0.8167	0.5278	0.8889
1.1332	15.0	750	2.5649	0.8650	0.8167	0.5278	0.8889
1.1332	16.0	800	2.6585	0.8557	0.7757	0.5	0.8889
1.1332	17.0	850	2.5906	0.8468	0.7476	0.4861	0.9028
1.1332	18.0	900	2.8373	0.7888	0.7023	0.4722	0.9028
1.1332	19.0	950	3.3194	0.8650	0.8167	0.5278	0.8889
0.3298	20.0	1000	3.0140	0.8008	0.7301	0.4861	0.9167

Framework versions

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