You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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=0load=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!
The text was updated successfully, but these errors were encountered:
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.
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
andtrain_steps=
0
load=
true
from saved model. What can be used here as aflag
?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!
The text was updated successfully, but these errors were encountered: