Hello, i have a problem with preparing the compressed tarball (*.tar.gz) of my MLCube config files. My folder structure looks like this: ``` medperf ??? cwdm ??? mlcube ? ??? mlcube.yaml ? ??? workspace ? ??? additional_files ? ? ??? flair_gen.pt ? ? ??? t1ce_gen.pt ? ? ??? t1_gen.pt ? ? ??? t2_gen.pt ? ??? parameters.yaml ??? project ??? ... ??? ... ``` My `mlcube.yaml` looks like this: ``` name: melide_GlobalSynthesis_BraTS2024 description: Submission to the BraTS 2024 Global Synthesis Challenge. Team 'melide'. Model generates missing modality MR images. authors: - {name: Team melide} platform: accelerator_count: 0 docker: # Image name image: docker.synapse.org/syn60334949/cwdm:v1.0 # Docker build context relative to $MLCUBE_ROOT. Default is `build`. build_context: "../project" # Docker file name within docker build context, default is `Dockerfile`. build_file: "Dockerfile" tasks: infer: # Computes predictions on input data parameters: inputs: { data_path: data/, parameters: parameters.yaml, parameters_t1gen: additional_files/t1_gen.pt, parameters_t1cegen: additional_files/t1ce_gen.pt, parameters_t2gen: additional_files/t2_gen.pt, parameters_flairgen: additional_files/flair_gen.pt, } outputs: {output_path: {type: directory, default: predictions}} ``` Following the instructions I first run: ``` ~/medperf/cwdm/mlcube/workspace$ tar -czf additional_files.tar.gz -C ~/medperf/cwdm/mlcube/workspace/additional_files . ``` When I then run: ``` ~/medperf$ python scripts/package-mlcube.py --mlcube cwdm/mlcube/ --mlcube-types 'model' --output melide.tar.gz ``` I get the following output: ``` Docker runner configuration identified Remember to push your image to a registry with: docker push docker.synapse.org/syn60334949/cwdm:v1.0 ERROR: Additional inputs must point to files in "additional_files" ``` Do you know how to solve this problem?

Created by irdeirfp
@irdeirfp , Yes, that looks correct to me.
@vchung Thanks, then I'll manually create the config tarball. Is this structure correct? ``` model.tar.gz ??? parameters.yaml ??? mlcube.yaml ??? additional_files.tar.gz ??? t1_gen.pt ??? t1ce_gen.pt ??? t2_gen.pt ??? flair_gen.pt ``` Thanks for your help.
@irdeirfp , Yes, I was going to suggest that it may be more convenient at this point to manually create the config tarball instead of trying to debug. Apologies for the inconvenience.
@vchung Would it also be possible to manually create a `*tar.gz` file similar to the on in your dummy model?
@vchung Thanks. This is what I actually tried in the first place, but I always get this `ERROR: Additional inputs must point to files in "additional_files"` error and don't really understand what the problem is, as my additional inputs actually do point to files in the `mlcube/workspace/additional_files` folder.
@irdeirfp , Great question. When using the convenience script provided by medperf (`scripts/package-mlcube.py`), you will not need to follow the instructions from the previous section, that is: the step where you had manually created the additional_files tarball. Instead, the script will recognize that the `addtional_files.tar.gz` will need to be created on your behalf based on your mlcube.yaml file. So for example, given this dummy model MLCube: ```text dummy_model ??? mlcube.yaml ??? workspace ??? additional_files ?   ??? foo.txt ??? parameters.yaml ``` After running the script, the resulting tarball will look like this: ```sh $ python scripts/package-mlcube.py --mlcube ~/Desktop/dummy_model --mlcube-types model --output ~/Desktop/model.tar.gz ...[truncated] ? Done! $ tar -tf ~/Desktop/model.tar.gz parameters.yaml mlcube.yaml additional_files.tar.gz ``` Hope this helps!

Error when preparing the compressed tarball (*.tar.gz) of MLCubes configs page is loading…