search

MIC-DKFZ / nnUNet

Apache License 2.0
5.79k stars 1.74k forks source link

How can i use pretrained model that was made by fine tunning #2005

Open copiglet opened 7 months ago

copiglet commented 7 months ago

I used this command nnUNetv2_train TARGET_DATASET CONFIG FOLD -pretrained_weights PATH_TO_CHECKPOINT

like this nnUNetv2_train 050 3d_lowres 0 -pretrained_weights /mnt/nas03/phenomx/cream_dental/Task040_PDT/nnUNet_results/Dataset100_TARGET/nnUNetTrainer__nnUNetPlans__3d_lowres/fold_all/checkpoint_best.pth

but after training, when i predicted this model it didn't make any of output.

and during training, the dice score showed only 0.0

As an additional word, Task040_PDT was SOURCE_DATASET and Dataset100_TARGET was TARGET_DATASET and 050 is the new TARGET_DATASET

the pretrained model that i used was made from them.

Can you tell me what should i do?

Do i have to set TARGET_PLANS_IDENTIFIER and SOURCE_PLANS_IDENTIFIER like when i did fine-tunning?

dojoh commented 5 months ago

hey, just to make sure. does it work if you do not use any pre-trained weights? did you follow the instructions here https://github.com/MIC-DKFZ/nnUNet/blob/master/documentation/pretraining_and_finetuning.md ?