What are steps to predict crystal systems for a custom (user defined) dataset

[Qais-Ali-5]

Thanks for nice work! I am little confused how may one use this for classifying their own dataset. For instance if I have a dataset of various 1D XRD patterns (be it measured or synthetic) with intensity vs 2theta information in folder. Few instructions are there for example;
It is mentioned that customed datasets can be evaluated by with Dataset= flag and train_steps= 0 load= true from saved model. What can be used here as a flag​ ?

Further if you may elaborate step by step how to use all the three models with 7 way crystal types to predict for a user defined dataset of XRD patterns?

Secondly does one need to first generate model? Or the pre-trained models can directly be used for prediction? Thanks in advance!

[AGI-init]

Dataset=<path.classname> should work. So for example, if you have a file called MyXRD.py with a Pytorch Dataset class called MyDataset, the command should look like: Dataset=MyXRD.MyDataset.

If your Dataset has train=<boolean> in its signature, it can toggle training vs. evaluation data. Then you can evaluate on the evaluation data just by adding train_steps=0 load=true. To select the model type, use task=NoPoolCNN or task=CNN or task=MLP.

Yes, one needs to generate (train) a model for load=true to work.