Murati, persistent Python and tips
7 minutes of Mira Murati
A great, great introduction to LLMs and transformers - and a chance to ogle at OpenAI’s CTO!
Covering attention, tokens, RLHF etc in 7 minutes enough for most.
Persistent Python
Nicholas Carlini’s great addition to Python DS stack - an environment state freezing machine!
Briefly, it’s a ~100 line script that allows me to snapshot the Python environment at an arbitrary point in time, and then interactively make changes to the code and run different experiments on the current state.
Paper release checklist
Again, Nicholas at it with a great checklist covering re-checking paper title to short scripts for e.g.
[x for x in open("/tmp/paper.tex") if 'TODO' in x or 'FIXME' in x]Tips, cheatsheets and sheets
An awesome list of tools, cheatsheets, awesome lists:
Generating generators + another article
You can iterate, or you can generate for:
a kind of lazy evaluation and memory savings (you produce the next item only when needed/requested)
producing an arbitrary number of items as opposed to generators like iterate()
(citing) automatically implementing __iter__(), __next__(), and StopIteration which otherwise, need to be explicitly specified
Coroutines
Another iteration type that can run asynchronously:
async def factorial(n):
if n == 0:
return 1
else:
return n * await factorial(n - 1)
async def main():
result = await factorial(10)
print(result)
asyncio.run(main())Welcome to Teleogenic/Boi Diaries❣️
Other places I cross-post to:
Reuse
Citation
@online{kogan2023,
author = {Kogan, Zakhar},
title = {Murati, Persistent {Python} and Tips},
date = {2023-09-17},
url = {https://teleogenic.com/posts/230910-murati/},
langid = {en}
}