ImperialCollegeLondon / covid19model

Code for modelling estimated deaths and cases for COVID19.
MIT License
944 stars 271 forks source link

Provide output as text files. #59

Closed g-rutter closed 4 years ago

g-rutter commented 4 years ago

Is your feature request related to a problem? Please describe. It's quite costly to obtain the output of this model at the moment. One has to prepare their system to run it, then do so for a few hours.

Describe the solution you'd like If you guys are running it most days anyway, couldn't you just provide the output? I see the graphs on your website, but no raw data output from anything I can see. It would be very helpful!

Describe alternatives you've considered Putting the graph into one of those graph digitisers. Or just running it myself.

Additional context Just a thank-you for your work on this!

becare-rocket commented 4 years ago

How long does it take to run? I'm running it now. Only took about 15 minutes to setup. What kind of machine are you using. I will benchmark mine when it finishes...

becare-rocket commented 4 years ago

Approximate benchmark is it takes about 40 minutes on an old mac-mini. I would bet it would run about 4x faster on a newer machine. I give this just so that other users running for the first time can have an idea on how much time it would take.

becare-rocket commented 4 years ago

G-Rutter -- I think the run takes about an 40 minutes on a macmini, although I am not sure I get the same results as they are presenting. To get the data in tabular form, find make-table.r and uncomment the commented lines that do a write.csv. You can change the target directory from "figures" to something else you prefer. I suggest setting up the batch job to run once around 1300Z so as to capture the current day's Euro CDC data. After that -- no need to resort to reading pixel's from a chart. If you are lucky you can get the Imperial College Team to do it for you, but it is not that hard. I wrote a small r function to translate their csv's back into the format that my R model expects transposing the rows and columns.

s-mishra commented 4 years ago

Hi @g-rutter sure we are working on it to provide csv or just a model that runs in like an hr or so. @becare-rocket that 40 min is the debug run not the full run.

jarib commented 4 years ago

@g-rutter I have a script here that generates a CSV per country from the model output: https://github.com/jarib/covid19model/blob/6e4a0e36e220333d43b8f6d75480b2e16cbb9240/make-country-csvs.r

@s-mishra Let me know if you'd like a pull request that drops this under web/ somewhere.

g-rutter commented 4 years ago

@s-mishra Thanks for your response. I am glad to hear that. I think the issue should remain open if your advice is "we are working on it" - that would be the usual approach.

@jarib Thanks for this. It should be helpful to the Imperial team.

payoto commented 4 years ago

Hello @jarib, Lovely feature! It was on my todo list, I'm gonna pull it into my fork on branch https://github.com/payoto/covid19model/tree/community-contribs , if you don't mind. I'm maintaining a branch of community features on my fork feel free to have a look for something of interest.

@s-mishra It would be nice to reopen this and tag it with the community label as we discussed the other day, similarly with PR #66 which seems related.