This repository holds the scripts used to train the ANNs used in the KinKal track fitting adapter for the Mu2e experiment
The first step to run the notebook is to install a kernel that includes support for everything listed in requirements.txt. For working on nersc, see https://docs.nersc.gov/services/jupyter/#conda-environments-as-kernels Then, you can load a particular notebook. Note that data is stored on NERSC in: /global/cfs/cdirs/m3712/Mu2e/TrkAna. On my macbook pro I used miniconda to install my kernel
TrainBkg.ipynb is used to separate true electron track hits from those generated by background particles TrainLRDrift.ipynb is used to separate hits which have accurate drift information, including left-right sign, from hits with either poor drift measurement (due to ion cluster effects) or an ambiguous left-right sign.
The first step to run the notebook is to install conda
(or conda
) and JupyterLab. I personally prefer conda
. Installation packages are available here.
Once you have installed Mamba, you can install JupyterLab with:
conda install -c conda-forge jupyterlab
Now, we need to create an environment containing the packages required by the tutorial. In the KKTrain
directory run:
conda create -n KKTrain -c conda-forge --file requirements.txt
This will create a KKTrain
environment. In order to access it you can run
conda activate KKTrain
We then want to add this environment to the list of kernels available to JupyterLab:
python -m ipykernel install --user --name KKTrain --display-name KKTrain
Now, go back to your base
environment and launch JupyterLab:
conda activate base
jupyter-lab
In order to use the KKTrain
environment you just created select "KKTrain" in the list of kernels in the upper right part of the interface. Then select one of the training scripts to run from the file browser.