segformer-b5-finetuned-magic-cards-230117-3

This model is a fine-tuned version of nvidia/mit-b5 on the andrewljohnson/magic_cards dataset. It achieves the following results on the evaluation set:

• Loss: 0.0691
• Mean Iou: 0.6585
• Mean Accuracy: 0.9878
• Overall Accuracy: 0.9912
• Accuracy Unlabeled: nan
• Accuracy Front: 0.9978
• Accuracy Back: 0.9777
• Iou Unlabeled: 0.0
• Iou Front: 0.9978
• Iou Back: 0.9777

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-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Front	Accuracy Back	Iou Unlabeled	Iou Front	Iou Back
1.2232	0.37	20	0.4691	0.6041	0.9201	0.9218	nan	0.9252	0.9150	0.0	0.9252	0.8870
0.2718	0.74	40	0.1983	0.6509	0.9764	0.9785	nan	0.9826	0.9702	0.0	0.9826	0.9702
0.255	1.11	60	0.0939	0.6524	0.9785	0.9794	nan	0.9812	0.9758	0.0	0.9812	0.9758
0.1103	1.48	80	0.0682	0.6536	0.9804	0.9813	nan	0.9830	0.9779	0.0	0.9830	0.9779
0.1373	1.85	100	0.1260	0.6631	0.9946	0.9961	nan	0.9989	0.9903	0.0	0.9989	0.9903
0.0566	2.22	120	0.1558	0.6578	0.9868	0.9912	nan	0.9999	0.9736	0.0	0.9999	0.9736
0.1535	2.59	140	0.1330	0.6558	0.9838	0.9883	nan	0.9973	0.9703	0.0	0.9973	0.9703
0.0586	2.96	160	0.2317	0.6599	0.9899	0.9933	nan	1.0000	0.9798	0.0	1.0000	0.9798
0.0727	3.33	180	0.1018	0.6586	0.9880	0.9919	nan	0.9995	0.9764	0.0	0.9995	0.9764
0.3588	3.7	200	0.1151	0.6608	0.9912	0.9939	nan	0.9993	0.9831	0.0	0.9993	0.9831
0.0463	4.07	220	0.0538	0.6610	0.9915	0.9934	nan	0.9969	0.9862	0.0	0.9969	0.9862
0.046	4.44	240	0.1201	0.6581	0.9871	0.9912	nan	0.9991	0.9751	0.0	0.9991	0.9751
0.0468	4.81	260	0.0691	0.6585	0.9878	0.9912	nan	0.9978	0.9777	0.0	0.9978	0.9777

Framework versions

• Transformers 4.25.1
• Pytorch 1.12.1
• Datasets 2.8.0
• Tokenizers 0.13.0.dev0