When converting a Caffe Model with CoreMLTools I found a Layer limitation that I'm trying to fix on my own fork of the project.
I've successfully ran:
cmake .
make
My problem now is how would I use this version I built in the normal workflow?
Try to replace the caffe subtree under coremltools/deps/caffe/ with your own adjusted caffe (Or merge the original CoreML caffe subtree with your changes).
Related
Using --profile makes shake generate a report.html file from which one can run queries and generate a dependency graph of build rules. Would it be possible to get that graph in graphviz format instead of needing a browser to display it?
If you run --profile=report.json then you get a JSON file containing all the data that goes into the profiling. From that, it should be quite easy to generate a GraphViz output. If there is demand, a GraphViz generation mode could be added to --profile.
However, past experience suggests that a GraphViz file of all dependencies is just too large to be practically viewed. To make the graph view practical you either need to group/filter the graph, or use a better graph viewer.
In Arango's Game of Thrones example dataset, they show a graph view where character nodes are labeled with two attributes: name and surname. How is this done?
Arango graph viewer example with Game of Thrones character data
The help doesn't explain it.
Graph Viewer Options
Nodes Options Menu:
Label (string): Nodes will be
labeled by this attribute. If node attribute is not found, no label will be displayed.
That makes it sound like only a single attribute can be used, yet their example shows the use of two attributes.
Luckily it's ArangoDB open source, so it's possible to achieve it and quite easily
here's quick and dirty commit adding such a functionality
to test it, not deploying it into production as it can cause unexpected security issues
clone on linux https://github.com/pavelsevcik/arangodb, cd into it, switch branch to feature/graph-viewer-config-labels--quick-and-dirty and run
cd js/apps/system/_admin/aardvark/APP
npm install
npm run grunt
mv node_modules ..
zip -r ../app.zip .
mv .. /node_modules .
then deploy created js/apps/system/_admin/aardvark/app.zip via services > add service > upload > set mount point to /aardvark
then change in url _admin/aardvark to aardvark and test it
here's Feature Request: Support of multiple attributes for displaying labels in Graph Viewer
The Graph Viewer does not support multiple attributes for displaying labels.
I took a screenshot of the graph without any labels and edited it in Paint.NET, that's the whole secret. This was for full control about the font face, size and placement. This combination was the easiest way to do it when I wrote the AQL tutorial.
Feel free to open a feature request on GitHub nonetheless.
currently looking through a boilerplate called react-redux-start-kit by Dave Zuko, and there is a folder called blueprints. Apparently it is for a library called redux-cli, and having no clue what that is I did some research. (Link to redux-cli)
The problem is, the documentation for redux-cli didn't really say what it is. I have also read that redux-cli makes it faster to build apps. Could someone please explain to me what redux-cli is, and how it works?
redux-cli is a utility that speeds up development by reducing work required to create basic elements of your app, e.g. components. Usually, when you create such component, you have to create file for a component, test suite and they all start with a certain structure that you usually copy / paste from other, existing components. With redux-cli you just use simplified commands in your console to auto-generate those basic versions of a new component and you can just jump in and start writing the essence of it.
blueprints are simply templates, redux-cli uses when running commands (for example when running command to create new component it will look at blueprints directory first and see if you have customized template for how you want your components to be initialized). Documentation is pretty clear I'd say - https://github.com/SpencerCDixon/redux-cli#creating-blueprints
I have a CEF-2454-based project that I wish to upgrade to 2623. However, I have made some changes to libcef needed for the project and I want to incorporate all changes made in 2623 without discarding my own changes. This raises me some questions:
What git merge strategy should I employ?
Can I build in my old 2454 directory, or I need to merge then build from scratch?
Should I merge 2526 then 2623, or I can directly merge 2623?
What is the proper way to incorporate a new CEF release into my own project?
Typical way is to merge your changes onto new CEF version. Unfortunately all other methods are more difficult.
You can try to take a diff between your changes and original CEF 2454, and try to apply that patch on 2623.
What's currently the best way to version multiple subprojects of one root project in Bazaar?
I'm familiar with bzr-externals and scmproj. I'm more drawn to bzr-externals, since it allows to use builtin commands (I'm using Bazaar Explorer), however it seems to lack some features I would like.
My project looks like this:
CommonLibrary
ProjectA (uses CommonLibrary)
ProjectB (uses CommonLibrary)
However, I would like to be able to store some project-level files (and have them versioned too), because I'm working in Visual Studio (which needs solution file containing information for all its projects).
In almost all cases, I would like both projects to share the same common library.
Also, currently I'm versioning them all under 1 branch - however when I choose any of the above solutions, I will need to split them, but I don't know how (I would like to preserve all history).
What's best way to accomplish this?
In order to split your branch, look at the bazaar split command. It does the following: Split a subdirectory of a tree into a separate tree.
I also mainly use bzr-externals because it uses the builtin commands.