A compute delegation package for the StochSS family of stochastic simulators
Project description
StochSS-Compute
StochSS-Compute is a compute delegation server for the StochSS family of stochastic simulation software. StochSS-Compute allows for one to run StochSS or GillesPy2 simulations on distributed cloud compute resources.
|
Installation
1. Installing dependencies & stochss_compute
PyPI Install:
pip install stochss_compute
Start up the server along with dask cluster:
stochss-compute-cluster
If you already have a dask cluster running on localhost:8786:
stochss-compute
Git Clone Install
Using a python virtual environment (documentation):
git clone https://github.com/StochSS/stochss-compute.git
cd stochss-compute
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
OR
Global install of dependencies:
git clone https://github.com/StochSS/stochss-compute.git
cd stochss-compute
pip install -r requirements.txt
2. Start up the server and compute backend
Using the startup script, which starts up a dask cluster, along with scheduler and workers:
python -m stochss_compute.launch cluster
If you already have a dask cluster running on localhost:8786:
python -m stochss_compute.launch
Docker
As an alternative to the above steps, you can use Docker.
Example Tutorial
1. Run this Docker command:
JUPYTER_PORT=7888 && \
docker run -it --rm \
-p $JUPYTER_PORT:$JUPYTER_PORT \
-p 8787:8787 \ # for dask dashboard
stochss/stochss-compute:examples \
jupyter notebook \
--port $JUPYTER_PORT
2. Open the link provided by the Jupyter Notebook server in your browser.
3. Open and run the self-contained Tutorial_1-Local.ipynb
Starting a server
docker run -it --rm --network host stochss/stochss-compute:latest
- Stochss-Compute is now running on localhost:29681.
- The cache defaults to the current working directory under
cache
. To set a new path for the cache, you can pass one todocker run
:
docker run -it --rm --network host -v $PWD/MyCache:/usr/src/app/cache stochss/stochss-compute
Cache Behavior
- Simulation results are given a unique identifier based upon the type of solver/algorithm, the model itself, and any other arguments passed to that simulation's
run()
call. - Results are stored on disk in json format.
- Results are 'anonymized', that is, variables and parameter names are converted to unique alphanumeric identifiers.
- Subsequent requests that match to cached results will automatically return the cached results.
- All other factors being the same, requests that differ only in the
number_of_trajectories
are associated with the same results object.
License
StochSS-Compute is licensed under the GNU General Public License version 3. Please see the file LICENSE for more information.
Acknowledgments
This work has been funded by National Institutes of Health (NIH) NIBIB Award No. 2R01EB014877-04A1.
StochSS-Compute uses numerous open-source packages, without which it would have been effectively impossible to develop this software with the resources we had. We want to acknowledge this debt. In alphabetical order, the packages are:
- Jupyter – web application for creating documents containing code, visualizations and narrative text
- Dask.Distributed – a library for distributed computation
Finally, we are grateful for institutional resources made available by the University of North Carolina at Asheville, the University of California at Santa Barbara, and Uppsala University.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for stochss_compute-1.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 20117172cf8eddab6188c5a83f35815b3121ef23f7f0d41ec8f004144c18e18c |
|
MD5 | 7b27f06ca394c1f98c2017d1ee42748d |
|
BLAKE2b-256 | 613936cd294bcc724c259c87a8564de510277640922ed15c7198608df9944746 |