Skip to content

Latest commit

 

History

History
52 lines (40 loc) · 1.42 KB

README.md

File metadata and controls

52 lines (40 loc) · 1.42 KB

Build Status codecov

How to use the package to reproduce the model and get predictions

  1. Create a conda environment, with rdkit, and install the package.
conda create -n qctime -c rdkit rdkit
conda activate qctime
cd models/qc_time_estimator
python setup.py install
# or for development
pip install -e .
  1. Train the model to generate and save the model's pkl file (used for prediction and tests)
python models/qc_time_estimator/qc_time_estimator/train_pipeline.py 
  1. Next, after generating the model, you can run tests (optional)
cd models/qc_time_estimator
pytest
  1. Finally, you can run predictions in Python:

Input file example:

nthreads,driver,method,restricted,cpu_clock_speed,cpu_launch_year
12,gradient,b3lyp,True,2500.0,2014
4,energy,wb97x-d,True,2100.0,2016
8,gradient,b3lyp,True,2200.0,2016
4,energy,pbe,True,2500.0,2014
2,energy,hf,True,2600.0,2013
4,energy,wb97x-d,True,2100.0,2016
16,energy,pbe,True,2100.0,2016
4,gradient,b3lyp,True,2500.0,2014
from qc_time_estimator.predict import make_prediction
import pandas as pd

test_input = pd.read_csv('path/to/file.csv')
predictions = make_prediction(input_data=test_input)