Python template Model_Inference.py -- Hard-coded material selection

[wkirgsn]

Hi,

we figured the python template Model_Inference.py is written such that it would need to be called 5 times together with editing the material selection line in order to receive all 5 material predictions.

Since the folder structure and data set paths are clear, do you mind if that file is rewritten such that it just loops over all 5 materials, producing all expected material prediction CSVs, and, hence, requires no editing?

[minjiechen]

You can keep it as 5 separate calls with materials at the calling index. Note you can develop different models for different materials as long as you can explain the reasons behind why.

We intent to create an "all-star" model by selecting the best models for each material across different teams.

[wkirgsn]

Hey, sorry, I have another problem with this template python file.
It assumes there is a "Testing" folder in the current working directory ( the same dir where the python template resides) from which data is loaded and in which estimates are stored.
However, in the provided zip, there is a "Result" folder instead.
What is the desired folder structure?

[wkirgsn]

Would the following structure be acceptable?

.
├── Model
│   ├── data
|   |     └──  input
|   |         └──  test
|   |               └── Testing
│   └── src
│       └── Model_Inference.py
└── Result
        ├── Volumetric_Loss_Material A.csv
        ├── Volumetric_Loss_Material B.csv
        ├── Volumetric_Loss_Material C.csv
        ├── Volumetric_Loss_Material D.csv
        └── Volumetric_Loss_Material E.csv

[haoranli1031]

Hello @wkirgsn, you may need to copy the Testing folder (the one we provided in the master folder, structure shown below) into the working directory during your test, which can be removed later and doesn't need to be included into the submission. This is the folder where datasets are loaded from.

.
2023 MagNet Challenge Testing Data (Public)
├── Training
│   ├── ...
└── Testing
        ├── Material A
        │   ├── B_Field.csv
        │   ├── Frequency.csv
        │   ├── Temperature.csv
        ├── Material B
        │   ├── ...
        ├── Material C
        │   ├── ...
        ├── Material D
        │   ├── ...
        ├── Material E
             ├── ...

On the other hand, the output files of predicted results are still stored in the Result folder.

[wkirgsn]

Ok, we would describe in a README, where exactly to place the Testing folder then.
I updated the proposed folder structure in the comment above, just to make it super clear what we imagine for the submission.

We will have the Model_Inference.py script edited to account for that folder structure of course