Skip to content

Instantly share code, notes, and snippets.

View lgastako's full-sized avatar

John lgastako

  • Francon & Heyer
  • Milky Way Galaxy, Third Rock from the Sun
View GitHub Profile
@sciolizer
sciolizer / puzzle.hs
Last active December 5, 2024 20:04
A haskell puzzle about representing functions with various arities.
-- Suppose we have the following expression type for a dynamically typed language
data Value
= VNull
| VInt Int
| VString String
| VNative NativeFunction
data NativeFunction = NativeFunction Int ([Value] -> IO Value)
arity :: NativeFunction -> Int

🤖 Agent Team 🤖

This example shows how to build an Agent having multiple agents as tools. This "Agent Team" works together to answer questions.

A more basic example can be found here

Install

pip install git+https://github.com/neuml/txtai
@getaaron
getaaron / irs-get-human.md
Created April 1, 2024 23:01
Get a person at the IRS
  • Call 1-800-829-1040
  • Press 1 for English (or other language as desired)
  • Press 2 for personal tax
  • Press 1 for form / tax history
  • Press 3 for other
  • Press 2 for other
  • Ignore 2 SSN prompts till you get secret other menu
  • Press 2 for personal tax
  • Press 3 for other
  • Wait for agent!
@anka-213
anka-213 / CheckStrictness.hs
Last active January 6, 2024 10:03
A type class for statically checking if data is fully strict
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FlexibleContexts #-}
module CheckStrictness where
import GHC.Generics
@moyix
moyix / CodeGen_GPTJ_Conversion.md
Last active May 5, 2025 14:22
How to convert the SalesForce CodeGen models to GPT-J

Using Linear Algebra to Convert a Large Code Model

Background

The SalesForce CodeGen models are a family of large language models trained on a large amount of natural language data and then fine-tuned on specialized datasets of code. Models of size 350M, 2B, 6B, and 16B parameters are provided in three flavors:

  • nl, the base model trained on The Pile, a large natural language dataset compiled by EleutherAI
  • multi, which is fine-tuned from the nl model on a dataset of code in multiple languages, scraped from GitHub, and
  • mono, which is fine-tuned from the multi model on Python code only.
@lgastako
lgastako / LambdaFizzBuzz.py
Last active July 11, 2023 13:30
Finally figured out how to write fizzbuzz in python
ADD = lambda a: lambda b: \
lambda f: lambda x: b(f)(a(f)(x))
MULT = lambda a: lambda b: \
lambda f: lambda x: a(b(f))(x)
ZERO = lambda f: lambda x: x
ONE = lambda f: lambda x: f(x)
TWO = ADD(ONE)(ONE)
THREE = ADD(ONE)(TWO)
FOUR = ADD(TWO)(TWO)
@Kazark
Kazark / CurryHoward.lhs
Last active June 6, 2024 23:47
Curry-Howard Tutorial in Literate Haskell
This is a tutorial on the Curry-Howard correspondence, or the correspondence
between logic and type theory, written by Keith Pinson, who is still a learner
on this subject. If you find an error, please let me know.
This is a Bird-style literate Haskell file. Everything is a comment by default.
Lines of actual code start with `>`. I recommend that you view it in an editor
that understands such things (e.g. Emacs with `haskell-mode`). References will
also be made to Scala, for programmers less familiar with Haskell.
We will need to turn on some language extensions. This is not an essay on good
{-# LANGUAGE NoImplicitPrelude #-}
{-# LANGUAGE OverloadedStrings #-}
module Fizz where
import Protolude
fizzBuzz :: IO ()
fizzBuzz = run fizzBuzzRules

Monads and delimited control are very closely related, so it isn’t too hard to understand them in terms of one another. From a monadic point of view, the big idea is that if you have the computation m >>= f, then f is m’s continuation. It’s the function that is called with m’s result to continue execution after m returns.

If you have a long chain of binds, the continuation is just the composition of all of them. So, for example, if you have

m >>= f >>= g >>= h

then the continuation of m is f >=> g >=> h. Likewise, the continuation of m >>= f is g >=> h.

@donovancrichton
donovancrichton / fibequiv.idr
Last active February 26, 2020 09:19
Proving extensionally equivalent fibonacci functions using Idris
%default total
||| A stream (an infinite list) of natural numbers representing the
||| fibbonacci sequence.
fibs : Stream Nat
fibs = f 0 1
where
f : Nat -> Nat -> Stream Nat
f a b = a :: f b (a + b)