Any plan to release the checkpoint?

[windygoo]

My reproduced results are {'Dice': tensor(0.7739, device='cuda:0'), 'Dice_Bal': tensor(0.6828, device='cuda:0'), 'F1': tensor(0.7739, device='cuda:0'), 'IOU_Jaccard': tensor(0.6311, device='cuda:0'), 'IOU_Jaccard_Bal': tensor(0.5344, device='cuda:0')}, which are lower than the reported results by a significant margin.

Dice: 0.7739 (reproduced) vs 0.7949 (in paper) IoU: 0.6311 (reproduced) vs 0.6596 (in paper)

Is any plan to release the checkpoint?

[jingweizhang-xyz]

The training log of this can be found at: https://wandb.ai/jingwezhang/sam_finetune_loss/reports/BCSS_fusion_focal_0125_iou_00625--Vmlldzo2MzMyMTk3 The model can be found at the end of this page.

[windygoo]

The training log of this can be found at: https://wandb.ai/jingwezhang/sam_finetune_loss/reports/BCSS_fusion_focal_0125_iou_00625--Vmlldzo2MzMyMTk3 The model can be found at the end of this page.

[jingweizhang-xyz]

Sorry, wrong link. https://api.wandb.ai/links/jingwezhang/2mov7snx

[windygoo]

Sorry, wrong link. https://api.wandb.ai/links/jingwezhang/2mov7snx

Get it, Thanks!

[windygoo]

Sorry, wrong link. https://api.wandb.ai/links/jingwezhang/2mov7snx

The checkpoint at the last epoch is saved? not the best on the validation set?