MLXRD
This is the opensource code for the PNNL LDRD project MDCRUST. In this code, Wwe apply a multi-task neural network to identify multi-compounds (Bastnaesite, Calcite, Re) from XRD (X-ray diffraction) data collected from a hydrothermal fluid system in synchrotron.
To run the code, Python3 and PyTorch are needed.
Folders:
dataPrepare:
- Data is the raw data, including theoretical and experiment dataset. Theoretical data is used for training, and experiment data is used for testing.
- DataPrepare.py is used to prepare dataset for training and testing from raw data.
train_test:
- DataLoader.py: load and preprocess the data.
- model_XRD.py: define neural network model.
- train_XRD: train model with training dataset and test.
- test_XRD: test data in trained model.
Run
To prepare data:
python DataPrepare.py
Train
python train_XRD.py
Test
python test_XRD.py
Contributors:
Yanfei Li, Xiaodong Zhao, Juejing Liu, Tong Geng, Ang Li, Xin Zhang.