abakus

8 ms·

31.
by abakus 7y ago
Very informative article! I use to work for some company that has a very strict no-merge, only rebase policy, because they want to preserve a straight-line commit history. But I think what they actually need is an algorithm that can draw &#
32.
by abakus 7y ago
https://github.com/blue-season/pywarm PyWarm is a lightweight, high-level neural network construction API for PyTorch. It enables defining all parts of NNs in the functional way.
33.
EfficientNet in 5 Minutes(blue-season.github.io)
3 points by abakus 7y ago|0 comments
34.
by abakus 7y ago
It is a nice font, but a bit too wide for my taste.
35.
Transformer Explained in 5 Minutes(blue-season.github.io)
3 points by abakus 7y ago|0 comments
36.
by abakus 7y ago
For a very easy to understand explanation of the transformer and attention mechansim, see: https://blue-season.github.io/transformer-in-5-minutes/
37.
by abakus 7y ago
I am looking for a starting programming language for my kids. I don't really like these lego-like drag-and-drop visual programming "languages" though. This definitely looks like in the right direction. Thanks for sharing!
38.
by abakus 7y ago
One way is via comments, like this: PyWarm, the functional API to build neural networks for PyTorch https://github.com/blue-season/pywarm
39.
Show HN: PyWarm, the functional API to build neural networks for PyTorch(github.com)
6 points by abakus 7y ago|0 comments
40.
by abakus 7y ago
Simple is better than complex. There should be one-- and preferably only one --obvious way to do it. https://xkcd.com/1987/
41.
by abakus 7y ago
Where did the dark energy come from? Before collapse these are just normal big stars, we cannot detect dark energy from them, but after explosion they suddenly become GEODE teemed with dark energy? I am a bit confused.
42.
Transformer in 5 Minutes(blue-season.github.io)
3 points by abakus 7y ago|0 comments
43.
Show HN: PyWarm, A cleaner way to build neural networks for PyTorch(github.com)
1 points by abakus 7y ago|0 comments
44.
Show HN: PyWarm, A cleaner way to build neural networks for PyTorch(github.com)
3 points by abakus 7y ago|0 comments
45.
by abakus 7y ago
Kudos to using toml for configuration.
46.
by abakus 7y ago
One step closer to the mars.
47.
by abakus 7y ago
YAML is horrible. Toml is much better. Even Json is not that messy.
48.
by abakus 7y ago
I find Julia's .> , .==, .*, ./ (dots for element-by-element ufunc)... really ugly. Numpy's design is cleaner and better.
49.
by abakus 7y ago
How about 1 space indentation? Just to be able to squeeze even more stuff into the 80 char limit? 0 space is a bit too extreme :)
50.
by abakus 7y ago
Just FYI: Huawei is one of the first companies that promoted the "996" schedule. https://github.com/996icu/996.ICU Huawei started the trend of laying off engineers once they reach 35 years old in China.
51.
by abakus 8y ago
Hey, but isn't TLO a ZERG player?
52.
by abakus 8y ago
Time to move forward. Python 2 is so 2010.
53.
by abakus 8y ago
But you still don't need Notebook, Cell evaluation is all you need, which can be done without all the notebook hassle.
54.
by abakus 8y ago
Just write python scripts instead in the first place. Import to Jupyter if you really, really need the notebook UI.
55.
by abakus 8y ago
10k loc is small enough to refactor by yourself in your not so busy work hours
56.
by abakus 8y ago
At the current state, openai five stands no chance against stronger teams (eg liquid lgd) of TI8.
57.
by abakus 8y ago
Ms word.
58.
by abakus 8y ago
This editor is probably gonna be banned in China if it got noticed by the government. Hell they banned Winnie the Pooh
59.
by abakus 8y ago
On the other hand, it follows the Numpy, Scipy's convention of using py as suffix instead of prefix.
60.
by abakus 8y ago
But this is not for man of industry!

More ›