mamba VS typing

> how the recall can grow unbounded with no tradeoff

this? https://github.com/state-spaces/mamba/issues/175

If you want to learn this stuff as a computer engineer, you can read the code here [0]. I find the math quite helpful.

[0]: https://github.com/state-spaces/mamba

This model was discussed recently: https://news.ycombinator.com/item?id=38522428 It's a new kind of ML model architecture that can be used instead of a transformer in LLMs.

See also the original repo from the paper: https://github.com/state-spaces/mamba

[2] - https://github.com/state-spaces/mamba

Out of curiosity, does anyone feel as though there's any benefit to linking to reddit when we can link to whatever the link is? I for one do not click the link and read discussion on reddit - if I wanted that sort of discussion, I would browse there, not HN.

The library is compatible with any auto-regressive model, not transformers. To prove our point we integrated Mamba, a new state-space model architecture, to the library. Try it out!

I ran the NanoGPT of Karparthy replacing Self-Attention with Mamba on his TinyShakespeare Dataset and within 5 minutes it started spitting out the following:

You might have come across the paper Mamba paper in the last days, which was the first attempt at scaling up state space models to 2.8B parameters to work on language data.

Iterable[str] is unfortunately evil as it matches str which is often unintended. (see: https://github.com/python/typing/issues/256) One would need both NOT-type and AND-type in order to properly handle these.

The basic motivation behind them is that functions can have arguments that can only take a specific set of values, and those functions return values/types change based on that input. Common examples are (you can find more here):

While yes 26 people liked the idea here: https://github.com/python/typing/issues/193

but looking at relevant issues on GitHub it seems this has been shot down repeatedly. python/typing#18, python/typing#213

See https://github.com/python/typing/issues/256 for a discussion.

Lower bounds are not supported for TypeVars, unfortunately.

Mypy allows that because initial versions of PEP-484 allowed that. This has changed; here's the current wording on the PEP:

> This is no longer the recommended behavior. Type checkers should move towards requiring the optional type to be made explicit.

https://www.python.org/dev/peps/pep-0484/#id29

Dunno, other, larger projects than the one I'm working on seem to run up against this from time to time. (rasa_core, to pick one example from near the top of a Google search; also Telethon, Blender, TensorFlow, Pandas. Guido also filed a bug on the typing module in an early version of Python 3.5 because of unexpected implications of this particular issue, so the problem isn't exactly purely theoretical.) That's aside from the wish for conceptual purity in the call signatures of classes and their subclasses, which is not always and automatically a bad wish to have; and the notion that a language that prides itself on its introspective faculties might want to make introspection of classes from the top of a class hierarchy possible, at least in theory? Perhaps to facility learning about the language and/or visualizing large class hierarchies easily, for instance?