News for hackers. Stuff that matters.

arogozhnikov

Re:Fastest Autograd in the West on Friday January 05, 2024 @05:51PM (#38886146)
> But these computation graphs are similar every time.

For now I'm testing one-after-another because failure of previous (overflow, too-small-transfer, etc) gives a hint about which part can be changed in the graph.

I've been thinking about vectorizing over multiple guesses at once, but don't see a fast reliable way to merge several graphs (also that induces significant burden on other parts of the project - autograd is just one component).

Re:Fastest Autograd in the West on Friday January 05, 2024 @02:20AM (#38876612)
Robotic planning of liquid handling. There are many potential alternative ways to achieve the same result, each resulting in its own computation graph.

I imagine any kind of trajectory planning for many agents faces similar challenges (e.g. robots in a factory).

Re:Ask HN: Are GitHub pages failing to deploy? on Thursday January 19, 2023 @02:46PM (#34444680)
same for me
sums it up pretty well
> API != User Interface. API = Programming Interface.

That's where we should start, but the other way around. Why do you keep trying to replace programming interfaces with bash calls?

Any example when calling cli compared to proposed alternative makes any difference to your spouse?

> then it should be pretty easy to test.

If it is that simple, you can just devote this to package like typer. That's what I point to - don't write it.

I see people somehow read GUI along the lines, while I never even considered that as an option. Added a note and please read completely before commenting

ok, I understand now my question got misinterpreted, I'll rewrite that.

CLI parsing (mapping to real calls) may have bugs. How do you test this?

Both options suggested eliminate this possibility completely.

> one of their core advantages is ease of automation and testing.

Any example when suggested options are any worse in this sense?

+1 for reading post
> Agree the CLI "shell" around the system should not have separate logic

that was the main point.

> should report back logic faults from within the library

And I suggest validating parameters passed to functions.

So what's this you disagree with?

It looks new - but is it any more complicated if after all that's what your call maps to?
It is easy only if you write CLI utility (or somewhat very light-weight).

Imagine you write ASR system (from CLI input to output - week(s) of training, requires several servers). Mind telling me more about using diff/stdin/stdout in this case?

Well, forget ASR, take any ML system.

Re:Gradient Boosting Explained in 3D on Wednesday July 27, 2016 @11:33AM (#12173632)
+1ed the issue. I was sure that's a problem on Safari side.