This repository packages the LumeLclsCuInjNn in lume_lcls_cu_inj_nn/model.py for execution with Prefect using the flow described in lume_lcls_cu_inj_nn/flow/flow.py using the variables:
| variable name | type | default |
|---|---|---|
| distgen:r_dist:sigma_xy:value | scalar | 0.413000000000000 |
| distgen:t_dist:length:value | scalar | 7.499772441611215 |
| distgen:total_charge:value | scalar | 250.000000000000000 |
| SOL1:solenoid_field_scale | scalar | 0.246000000000000 |
| CQ01:b1_gradient | scalar | -0.007400000000000 |
| SQ01:b1_gradient | scalar | -0.007400000000000 |
| L0A_phase:dtheta0_deg | scalar | -8.899700000000000 |
| L0A_scale:voltage | scalar | 70000000.000000000000000 |
| end_mean_z | scalar | 4.614700200000000 |
| variable_name | type |
|---|---|
| end_n_particle | scalar |
| end_mean_gamma | scalar |
| end_sigma_gamma | scalar |
| end_mean_x | scalar |
| end_mean_y | scalar |
| end_norm_emit_x | scalar |
| end_norm_emit_y | scalar |
| end_norm_emit_z | scalar |
| end_sigma_x | scalar |
| end_sigma_y | scalar |
| end_sigma_z | scalar |
| end_mean_px | scalar |
| end_mean_py | scalar |
| end_mean_pz | scalar |
| end_sigma_px | scalar |
| end_sigma_py | scalar |
| end_sigma_pz | scalar |
| end_higher_order_energy_spread | scalar |
| end_cov_x__px | scalar |
| end_cov_y__py | scalar |
| end_cov_z__pz | scalar |
| x:y | image |
| out_ymax | scalar |
| out_xmax | scalar |
| out_ymin | scalar |
| out_xmin | scalar |
This package may be installed using pip:
pip install git+https://github.com/jacquelinegarrahan/lume-lcls-cu-inj-nn
Install dev environment:
conda env create -f dev-environment.yml
Activate your environment:
conda activate lume-lcls-cu-inj-nn-dev
Install package:
pip install -e .
Tests can be executed from the root directory using:
pytest .
A script (epics_queue.py) has been packaged in the root of this repository that may use the LUME-services tools to continually queue model execution on EPICS PV monitors.
The environment for the LUME-services docker-compose system should be configured as described in the LUME-services demo.
Launch the compose application using the development environment packaged with this repository and configured as described above:
conda activate lume-lcls-cu-inj-nn-dev
lume-services docker start-services
In another window, run the model and deployment registration blocks using the notebook in examples/Run.ipynb.
conda activate lume-lcls-cu-inj-nn-dev
source demo.env
jupyter notebook Run.ipynb
In yet another window, configure the EPICS environment using a port forwarding from prod machines.
export EPICS_CA_NAME_SERVER_PORT=24666
ssh -fN -L $EPICS_CA_NAME_SERVER_PORT:<LCLS_PROD_HOST> $SLAC_USERNAME@$SLAC_MACHINE
export EPICS_CA_NAME_SERVERS=localhost:$EPICS_CA_NAME_SERVER_PORT
Where EPICS_CA_NAME_SERVER_PORT may be any available port, not necessarily 24666. The other variables should be provided by someone with context to SLAC systems.
conda activate lume-lcls-cu-inj-nn-dev
source demo.env
python epics_queue.py 1 1
Where the first argument 1 corresponds to the registered model_id and the second 1 corresponds to the registered deployment_id.
Once finished, you'll have to kill the port-forwarding process manually by finding the pid:
ps aux | grep ssh
kill <pid of your ssh process>
Use the dev on prod EPICS configuration:
conda activate lume-lcls-cu-inj-nn-dev
python epics_queue.py
Please follow/test above all steps before reproducing these here -
- Open a Double SSH Terminal with 2 Ports for Prefect UI and Server. Below example will start Lume Services on localhost:4204 and Prefect UI on localhost:4113
ssh -L 4113:localhost:8992 -L 4203:localhost:8993 thakur12@centos7.slac.stanford.edu -t ssh -L 8993:localhost:4203 -L 8992:localhost:4112 thakur12@PC101221
- Clone Lume Services Repo
- Install Lume Services
pip install git+https://github.com/slaclab/lume-services.git
- Run below code to make sure Lume Services has starte. You should get a prompt - All Services have started and are passing health checks.
cd lume-services/
conda env create -f dev-environment.yml
conda activate lume-services-dev
source docs/examples/demo.env
export STANFORD_USERNAME=<your_stanford_username>
export SCR_PAT=<your PAT from step 9>
lume-services docker start-services
- Open a Double SSH Terminal with a Port to open Jupyter Notebook. Below example will allow you to start a Jupyter notebook on localhost:8990
ssh -L 8990:localhost:8991 thakur12@centos7.slac.stanford.edu -t ssh -L 8991:localhost:1234 thakur12@PC101221
- Clone NN Injector Repo
- Follow the same steps as above - Install, Create Environment, Set Env Variables
cd lume-lcls-cu-inj-nn/
conda env create-f dev-environment.yml
conda activate lume-lcls-cu-inj-nn-dev
source demo.env
jupyter notebook --no-browser --port=1234
- Feel free to run Run.ipynb for a demo run or you can run the service that listens to epics/processes in batch -
export EPICS_CA_NAME_SERVER_PORT=24800
ssh -fN -L $EPICS_CA_NAME_SERVER_PORT:lcls-prod01.slac.stanford.edu:5068 thakur12@centos7.slac.stanford.edu
export EPICS_CA_NAME_SERVERS=localhost:$EPICS_CA_NAME_SERVER_PORT
python epics_queue.py 1 1
This README was automatically generated using the template defined in https://github.com/slaclab/lume-services-model-template with the following configuration:
{
"author": "Jacqueline Garrahan",
"email": "jacquelinegarrahan@gmail.com",
"github_username": "jacquelinegarrahan",
"github_url": "https://github.com/slaclab/lume-lcls-cu-inj-nn",
"project_name": "LUME LCLS cu inj nn",
"project_slug": "lume-lcls-cu-inj-nn",
"package": "lume_lcls_cu_inj_nn",
"model_class": "LumeLclsCuInjNn"
}