segformer-b5-finetuned-magic-cards-230117

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: 0.2096
• Mean Iou: 0.6629
• Mean Accuracy: 0.9944
• Overall Accuracy: 0.9944
• Accuracy Unlabeled: nan
• Accuracy Front: 0.9997
• Accuracy Back: 0.9891
• Iou Unlabeled: 0.0
• Iou Front: 0.9997
• Iou Back: 0.9891

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: 10

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
0.496	0.74	20	0.4441	0.6552	0.9838	0.9838	nan	0.9786	0.9890	0.0	0.9786	0.9869
0.1693	1.48	40	0.4098	0.6597	0.9897	0.9897	nan	0.9943	0.9851	0.0	0.9943	0.9849
0.1172	2.22	60	0.2734	0.6582	0.9874	0.9874	nan	0.9977	0.9770	0.0	0.9977	0.9770
0.1335	2.96	80	0.2637	0.6609	0.9914	0.9914	nan	0.9959	0.9869	0.0	0.9959	0.9869
0.0781	3.7	100	0.5178	0.6644	0.9966	0.9966	nan	0.9998	0.9933	0.0	0.9998	0.9933
0.1302	4.44	120	0.2753	0.6652	0.9978	0.9978	nan	0.9993	0.9962	0.0	0.9993	0.9962
0.0688	5.19	140	0.1458	0.6618	0.9926	0.9926	nan	0.9950	0.9903	0.0	0.9950	0.9903
0.0866	5.93	160	0.1763	0.6636	0.9954	0.9954	nan	0.9962	0.9946	0.0	0.9962	0.9946
0.0525	6.67	180	0.1812	0.6627	0.9941	0.9941	nan	0.9988	0.9895	0.0	0.9988	0.9895
0.0679	7.41	200	0.2246	0.6625	0.9937	0.9937	nan	0.9990	0.9884	0.0	0.9990	0.9884
0.0424	8.15	220	0.2079	0.6623	0.9934	0.9935	nan	0.9996	0.9873	0.0	0.9996	0.9873
0.0349	8.89	240	0.1559	0.6626	0.9939	0.9940	nan	0.9987	0.9892	0.0	0.9987	0.9892
0.0357	9.63	260	0.2096	0.6629	0.9944	0.9944	nan	0.9997	0.9891	0.0	0.9997	0.9891

Framework versions

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