lcls-injector-epics

An example implementation of LUME-EPICS and LUME-Model with the LCLS injector surrogate.

Run

Firstly, create the environment defined in environment.yml. Then in a linux shell, run the following command to start the EPICS server:

python server.py --no-calibration

If you want to use a client to visualize the results of the model in real time, open a second linux shell and run the following command to pull up a bokeh dashboard in your browser:

bokeh serve client.py --show

Notes

• currently the environment.yml uses a pip installed version of lume-epics from a local fork. Once the PR is accepted and the changes pulled into the main repo, this can be updated to use the conda version of the library
• Remember to change the serve flag and the pvname parameters in configs/epics_config.yml if you want to run on the live machine. Otherwise, new PVs will be created and served rather than reading from the live machine.
• Using the --calibration flag instead of --no-calibration will allow force the model to use decoupled input and output calibration parameters defined in configs/calibration.json.
• compound PVs (e.g. CAMR:IN20:186:R_DIST, a product of "CAMR:IN20:186:R_DIST" and "CAMR:IN20:186:R_DIST") can be defined by overriding the _prepare_inputs() method of lume_model.torch.PyTorchModel.
  • it may be best in future to create a new Variable class that allows this behaviour