241–250 of 617 posts
I built audiomass ( https://audiomass.co ). it is not a dev tool but a general productivity tool (allow people to record, edit and manipulate audio directly in the browser). I just wanted something that works as promised and respects the user (open source, 70kb total payload size, no ads, no tracking, feature complete, etc)
In addition to that I built a few other tools, like my link shortener. It does everything I need it to do the way I want it, but nothing more.
https://insightrend.com/?page=term-trending - like google trends but better
and https://ideasfilter.com/ a list of apps that you can copy and make money of
I built it because I was making screencasts and cutting out silence + mistakes was 90% of my editing time. This makes it a ton faster. There were some command-line scripts to do a similar thing but I wanted something visual with a fast feedback loop, where I could quickly preview how it would sound and tweak the parameters in real time.
I didn’t know anything about video and had never built anything with Swift before so it’s been a fun way to learn a bunch of new stuff.
* the core idea is to selectively use training instances to learn a compact versions of a model. Multiple models are produced with different tradeoffs between accuracy and size. The user is then free to pick one that fits his needs. See [2], [3].
* one can also find the "minimal dataset" a models needs to reach a certain accuracy - [4].
Currently working on:
* I am reformulating the problem to work with (twice) differentiable losses to speed-up the compaction process.
* Also, I am looking multiple "stages" of compaction to see if that leads to models with further reduced sizes.
Refs:
[1] https://compactem.readthedocs.io/en/latest/
[2] see how the model's generalization is affected, also has accuracy vs size curves: https://compactem.readthedocs.io/en/latest/key_ideas.html#wh...
[3] Compaction profile - a plot to visualize the range of size reductions between the original and compacted models: https://compactem.readthedocs.io/en/latest/usage.html#compac...
[4] https://compactem.readthedocs.io/en/latest/usage.html#optima...
I also developed a Lua debugger [2] and serializer [3], as well as took on maintenance of wxlua project for the last 5 years [4]; all open source.
[1] https://studio.zerobrane.com [2] https://github.com/pkulchenko/MobDebug [3] https://github.com/pkulchenko/serpent [4] https://github.com/pkulchenko/wxlua
https://github.com/agentultra/DataVersion — type-directed data migration library. I needed to migrate some data in a document database that had no schema with years of data in it to a format that had a schema. I leveraged some other stuff to achieve it but this library was a key piece to being confident that the transformations were correct and complete. Migrated the database and backfill it with a Lambda function using this library.
Those are just the most recent ones.
Have posted about this before, but I have been working on my Python library called "compactem". A way to make traditional ML models smaller: [1]. * the core idea is to selectively use training instances to learn a compact versions of a model. Multiple models are produced with different tradeoffs between accuracy and size. The user is then free to pick one that fits his needs. See [2], [3]. * one can also find the "mi…