tt-acm / colibri.grasshopper Goto Github PK

ask user if want to save file before run all iterations

Change the naming / description on the selector to one single domain, but let code remain the same.

replace them with " " (space)

check execute order: capture Rhino view after all Display_type components

Use a nice grey phong if no mesh colors, otherwise use mesh colors with a lambert. I can help!

when a slider set to min 0 to max 1 integer, Iterator has a problem with this.

Hello,

It would be really helpful for ML applications to add the option of exporting the whole design space (i.e. the csv file of all the different combinations of input model parameters, without the results) to a .csv file.

So far in my parametric models I have been achieving this using the fly component along with native GH components to save and export to excel. It'd be neat to have this within one component.

Kind regards,
Theodore.

Can we make steps have default values when nothing is connected?

Iterator skip over the step (2) when a slider was set to 1 to 4 integer, and no setting on Steps.

"Study Folder is not empty, do you want to override everything inside!"
1 - prompts you to choose yes or no for every single iteration of colibri
2 - (less important) grammar, should be "Study Folder is not empty! Do you want to override everything inside?"

When the Iterator has run once, and a new Parameter is added to the Parameters node running into the Phenome input of the Aggregator, and the Iterator is run again, the .csv file has the new column of data, but the headers are offset:

"img" becomes the header of the new parameter
"threeD" becomes the header of the images
no header of the .json column

this causes the folder to fail to load in design explorer

Iterator can not re-check the connected sliders' name when it is changed, unless reconnect it.

Firstly I'd like to say thanks for this cool project! I feel it's going to be very helpful and is coming at the right time considering how parametric models are becoming more and more popular.

I will try and add a few wishes mainly based on my latest on-going project, connecting Machine Learning and parametric modelling, through HB/LB/GH/DesignExplorer. In order to make things easier for responding to these I will split them up in separate issues. I hope that is ok, let me know if you want an aggregated post.

The first one, which is quite related to #10 is to add an option for Colibri to run a percentage of the design space, set by the user.

While selecting a number of iterations to run as a means of testing the sanity of the definition is fine (see #10), a percentage of simulations is something you would want to select in production models and stages. It is also critical for ML applications, where you only select a part of your design space to run.

Furthermore, there are two things that this has to achieve:

• random selection of design instances from the whole design space
• stratified selection across model parameters (i.e. as much as possible select equal instances of different parameters), in order to reduce the bias in the training set.

That's all for this one! Thanks!

Kind regards,
Theodore

Hello,

First of all congratulations on the release, I've been testing Colibri and it's quite nice!

As you might have witnessed, during parametric runs a lot of things can happen that might disrupt the simulation. In that case we are left with a .csv file that is incomplete.

Is it possible to restart the run, while Colibri reads the already existing data.csv and skips the runs previously completed?

Thanks in advance!

Kind regards,
Theodore.

I have discovered a fairly large bug in Colibri that is currently making it difficult to use. Namely, the names of Inputs from the Iterator do not match the actual data:

It seems to be a fairly simple bug to fix (just a list management issue) and it would stop me from having to manually edit the output CSV file after running Colibri. Here's a GH file that shows the issue:
https://www.dropbox.com/s/os3u9yagzstd5m9/colibriBug.gh?dl=0

-Chris

See this thread on the GH forum: http://www.grasshopper3d.com/forum/topics/colibri-aggregator-error-from-imgsetting?groupUrl=corestudio

Mark said he saw it go to twenty.

Iterator takes Slider, Panel, ValueList for next release.

to make the "tt-acm/Colibri.Grasshopper" as "root parent" repo, not forked from "bhowes-tt/Colibri.Grasshopper".

1. we can ask Github staff to extract this repo from "bhowes-tt/Colibri.Grasshopper".
2. transfer "Colibri.Grasshopper' ownership
3. or switch the relationship: let "bhowes-tt/Colibri.Grasshopper" fork from "tt-acm/Colibri.Grasshopper"

Hello,

The current work flow I am using incorporating supervised learning to parametric models consists of exporting an initial csv input file of the whole design space (see #12), exporting a .csv result file of a percentage of the design space (see #11), applying supervised learning that generates the .csv result file for the whole design space.

Given the special format of DesignExplorer it would be nice if there was an option to automatically change an input .csv to (and fro) a DesignExplorer format. Ofc this is all too easy manually but why not automate it!

Kind regards,
Theodore.

From #1 - changing the Steps values does not update the number of iterations that the MessageBox reports to the user. This happens after telling colibri to fly, then pressing no, then changing Steps values, then telling colibri to fly again.

need to take care of user's inputs with units

Check if folder existed, if not, create one.

Randomly pick 3 or 5 iterations for testing run to check if everything works fine.