Skip to content

Latest commit

 

History

History
40 lines (26 loc) · 1.2 KB

File metadata and controls

40 lines (26 loc) · 1.2 KB

River:

A visualization frontend for MCMC samplers

The visualization relies on some python libraries. Install them with

sudo pip install Flask Flask-Sockets Gunicorn

Run the server with

gunicorn -k flask_sockets.worker server:app

or if you need more options

gunicorn -k flask_sockets.worker server:app --bind 0.0.0.0:8000 --timeout 120 --log-level debug

server.py is a repeater that reads from a websocket. You can create samples using any system. As an example to create samples using wssh

python examples/test.py | wssh localhost:8000/sampler_input

Then open the client.html in a web browser

If you have pymc3 installed, try out some of the other examples

JSON visualization grammar

The frontend accepts JSON with the following format

format = "{" draw_types* "}"
draw_types = line | histogram | total | current
line = "\"linreg\": [" tuple* "]" 
tuple = "{\"x\":" number "} , {\"y\":" number "}"
total = "\"total_samples\":" number
current = "\"current_sample\":" number
histogram = "\"rvars\": [" vars* "]" 
vars = string ":" "[" samples "]"
samples = number*

Expect this grammar to be extended and made more modular