segformer-b5-miic-tl

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

• Loss: 0.2247
• Mean Iou: 0.4565
• Mean Accuracy: 0.9129
• Overall Accuracy: 0.9129
• Accuracy Unlabeled: nan
• Accuracy Circuit: 0.9129
• Iou Unlabeled: 0.0
• Iou Circuit: 0.9129
• Dice Coefficient: 0.8406

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Circuit	Iou Unlabeled	Iou Circuit	Dice Coefficient
0.2801	3.12	250	0.2305	0.4832	0.9663	0.9663	nan	0.9663	0.0	0.9663	0.8527
0.2785	6.25	500	0.2715	0.4800	0.9601	0.9601	nan	0.9601	0.0	0.9601	0.8511
0.208	9.38	750	0.2681	0.4811	0.9622	0.9622	nan	0.9622	0.0	0.9622	0.8538
0.2042	12.5	1000	0.2959	0.4650	0.9299	0.9299	nan	0.9299	0.0	0.9299	0.7879
0.1649	15.62	1250	0.2407	0.4340	0.8679	0.8679	nan	0.8679	0.0	0.8679	0.8150
0.1353	18.75	1500	0.2530	0.4543	0.9085	0.9085	nan	0.9085	0.0	0.9085	0.8336
0.126	21.88	1750	0.4934	0.4559	0.9119	0.9119	nan	0.9119	0.0	0.9119	0.7678
0.1196	25.0	2000	0.2896	0.4604	0.9209	0.9209	nan	0.9209	0.0	0.9209	0.7807
0.1149	28.12	2250	0.2210	0.4634	0.9268	0.9268	nan	0.9268	0.0	0.9268	0.8470
0.1095	31.25	2500	0.2215	0.4534	0.9067	0.9067	nan	0.9067	0.0	0.9067	0.8380
0.109	34.38	2750	0.2256	0.4243	0.8487	0.8487	nan	0.8487	0.0	0.8487	0.8077
0.1062	37.5	3000	0.2172	0.4497	0.8994	0.8994	nan	0.8994	0.0	0.8994	0.8363
0.1046	40.62	3250	0.2401	0.4551	0.9102	0.9102	nan	0.9102	0.0	0.9102	0.8387
0.1096	43.75	3500	0.2157	0.4582	0.9164	0.9164	nan	0.9164	0.0	0.9164	0.8425
0.1014	46.88	3750	0.2344	0.4573	0.9146	0.9146	nan	0.9146	0.0	0.9146	0.8411
0.1036	50.0	4000	0.2247	0.4565	0.9129	0.9129	nan	0.9129	0.0	0.9129	0.8406

Framework versions

• Transformers 4.36.2
• Pytorch 1.11.0+cu115
• Datasets 2.15.0
• Tokenizers 0.15.0