search

usnistgov / AusteniteCalculator

Software and GUI for uncertainty quantification of austenite phase fraction
Other
1 stars 0 forks source link

Rerunning Austenite Calculator with different files leads to the same result #10

Closed creuzige closed 2 years ago

creuzige commented 2 years ago

I tested the different cif files in Example 05 to see how much the composition variation would affect the phase fraction. Unfortunately, it looks like some of the links or information persist after the first analysis run. I ran the nominal composition cif files, as well as the high and low values. Each time the R_calc column would remain the same as the first file I ran. Rebuilding the docker image seemed to clear this initially, but now it looks like it's using a cached copy even when copying the directory, and then isn't changing.

=> CACHED [ 8/11] RUN mkdir /root/AustCalc 0.0s => CACHED [ 9/11] COPY ./ /root/AustCalc/ 0.0s => CACHED [10/11] RUN /root/g2full/bin/pip install -r /root/AustCalc/requirements.txt 0.0s => CACHED [11/11] WORKDIR /root/AustCalc/app/

dnewton600 commented 2 years ago

Sorry, this seems to be related to hard coding the austenite and ferrite file names (one of the other github issues). Tomorrow we can discuss in more detail what types of phases and file names to expect from the user's end, then things should be resolved when we fix this. Right now the app specifically looks for 'austenite-duplex.cif', for example. I think when we were first starting I wasn't sure how general the phase files needed to be.

creuzige commented 2 years ago

Sounds like a good plan. I agree that's likely the underlying issue...

dnewton600 commented 2 years ago

I think it's because I initially put in

phase_austenite = get_phase(data_path_wrap("austenite-Duplex.cif"), "Austenite", gpx) phase_ferrite = get_phase(data_path_wrap("ferrite-Duplex.cif"), "Ferrite", gpx)

which ignores the specific cif files that the user uploads :(

creuzige commented 2 years ago

I guess my recollection was that if you upload the files from dash, then the files get renamed as what's you've described (something else we can investigate/document). But I don't think we have a 'reset folder' button, so anything that gets placed there persists past the first analysis.

I've also changed the usage to point to an example folder for default files, which we also didn't really discuss how to implement before I took a stab at it :)

creuzige commented 2 years ago

Looks like this got resolved when we stopped hard coding things in. Thanks!