Issue exporting trained nnUNet model

[skim2257]

Unable to export the model zipfile to share the model.

I ran:

nnUNetv2_export_model_to_zip 
-o ../model.zip 
-c 3d_fullres 
-d 001 
-f 0 
-chk checkpoint_best.pth

This was the output:

Configuration 3d_fullres
Exporting fold_0
Traceback (most recent call last):
  File "miniconda3/envs/nnunet/bin/nnUNetv2_export_model_to_zip", line 33, in <module>
    sys.exit(load_entry_point('nnunetv2', 'console_scripts', 'nnUNetv2_export_model_to_zip')())
  File "nnUNetv2/nnunetv2/model_sharing/entry_points.py", line 59, in export_pretrained_model_entry
    export_pretrained_model(dataset_name_or_id=args.d, output_file=args.o, configurations=args.c, trainer=args.tr,
  File "nnUNetv2/nnunetv2/model_sharing/model_export.py", line 62, in export_pretrained_model
    zipf.write(source_file, os.path.relpath(source_file, nnUNet_results))
  File "miniconda3/envs/nnunet/lib/python3.10/zipfile.py", line 1737, in write
    zinfo = ZipInfo.from_file(filename, arcname,
  File "miniconda3/envs/nnunet/lib/python3.10/zipfile.py", line 502, in from_file
    st = os.stat(filename)
FileNotFoundError: [Errno 2] No such file or directory 'Dataset001_DATASET/nnUNetTrainer__nnUNetPlans__3d_fullres/fold_0/validation/summary.json'

Please excuse the sanitized paths in the output log. 
My `fold_0` folder looks like this:

-rw-r--r-- 1 sejinkim bhklab 236M May 18 2023 checkpoint_best.pth
-rw-r--r-- 1 sejinkim bhklab 236M May 18 2023 checkpoint_latest.pth
-rw-r--r-- 1 sejinkim bhklab 17K May 17 2023 debug.json
-rw-r--r-- 1 sejinkim bhklab 628K May 18 2023 progress.png
-rw-r--r-- 1 sejinkim bhklab 58K May 18 2023 training_log_2023_5_17_15_29_35.txt

I'm also happy to produce the Zipfile myself, if you can direct me to what `install_model_from_zip_file` expects. 

Thanks,
Sejin

[constantinulrich]

Hi, this function wants to zip the summary.json file as well which will be generated after the training and validation.
2 Possibilities: You could just run the missing validation (--val at the end of your training command) or just un comment the part here of the function.

Edid: you would need to adapt the function also in some other parts, should still be easy to do.
Best
Constantin