Henk Poley HenkPoley

Credit: How to write ML papers by Neel Nanda

You are chatbot that gives constructive analysis of the following work. Specifically, you care about the following criteria:

## Core Narrative Quality
- **Clear Claims**: Contains 1-3 specific, concrete claims that fit within a cohesive theme
- **Strong Motivation**: Clearly explains why readers should care ("so what?")

Disclaimer: I'm in the Top 1% of StackOverflow contributors with 23,315 rep points.

I asked 1 high-quality question in 2024, and it was closed almost immediately, and I haven't engaged with the site since.

If someone with 20,000+ karma has their nicely-formatted questions closed so quickly, what must the newbies and rank-in-file encounter? This is probably a big reason why it's declining.

Which GGUF is right for me? (Opinionated)

Good question! I am collecting human data on how quantization affects outputs. See here for more information: ggml-org/llama.cpp#5962

In the meantime, use the largest that fully fits in your GPU. If you can comfortably fit Q4_K_S, try using a model with more parameters.

llama.cpp feature matrix

See the wiki upstream: https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix

Model	Average	AGIEval	GPT4All	TruthfulQA	Bigbench
mlabonne/OmniTruthyBeagle-7B-v0 📄	57.8	45.72	77.49	76.16	50.18
mlabonne/NeuralOmniBeagle-7B-v2 📄	57.75	45.86	77.31	75.34	50.09
mlabonne/OmniBeagle-7B 📄	57.72	45.64	77.48	75.03	50.03
mlabonne/NeuralOmniBeagle-7B 📄	57.71	45.85	77.26	76.06	50.03
mlabonne/NeuralOmni-7B [📄](https://gist.github.com/mlabonne/4b5ecee86d0fd3714ba0cbd

Using Git Commit Message Templates to Write Better Commit Messages

Introduction

In this document, we will discuss how to set up and use Git commit message templates to write better commit messages.

Prerequisites

Git installed on your local machine
Basic understanding of Git and commit messages

How to Defang macOS System Protections

If you want to change things on the root drive of a Mac you will need to take some steps to disable the built in security of the system. Most of these steps are the same regardless if you are on Intel or Apple Silicon. If there is a difference it is noted.

Note that all of these things put a Mac into an unsupported and less secure state.

Make sure you either perform these steps in a VM or that you reset the protections after you are done poking around

Protections and Terms

(This list is not exahustive on the details of each. Check the links at the end for more info.)

	Benchmark is from an AI Cloud Rig:
	https://cloud.vast.ai/?ref_id=127244


	hashcat (v6.2.6-851-g6716447df) starting in benchmark mode
	CUDA API (CUDA 12.8)
	====================
	* Device #1: NVIDIA GeForce RTX 5090, 31610/32120 MB, 170MCU
	* Device #2: NVIDIA GeForce RTX 5090, 31610/32120 MB, 170MCU
	* Device #3: NVIDIA GeForce RTX 5090, 31610/32120 MB, 170MCU

	import json
	import random
	import mlx.optimizers as optim
	import mlx.core as mx
	import mlx.nn as nn
	import numpy as np

	from tqdm import tqdm
	import time
	from datetime import datetime

	@echo off
	setlocal enabledelayedexpansion

	rem \| Lines beginning with 'rem' are comments

	rem -------------------------------------------------------------------------------------------

	rem \| "Icon Upscaler" Script by ThioJoe: https://github.com/ThioJoe
	rem \| I created this script specifically for upscaling very low res (such as 256x256) icons like those in Windows. It uses a combination of Image Magick (to first improve the transparency edges) and the RealCugan-ncnn-vulkan upscaler which seems to perform the best for this use case. Of course, the script can be used for all sorts of images, but I have found this to yield the best results of any other method, and better than just using an upscaler alone.

	// You will function as a JSON api.
	// The user will feed you valid JSON and you will return valid JSON, do not add any extra characters to the output that would make your output invalid JSON.

	// The end of this system message will contain a typescript file that exports 5 types:
	// Prompt - String literal will use double curly braces to denote a variable.
	// Input - The data the user feeds you must strictly match this type.
	// Output - The data you return to the user must strictly match this type.
	// Errors - A union type that you will classify any errors you encounter into.
	// Tools - If you do not know the answer, Do not make anything up, Use a tool. To use a tool pick one from the Tools union and print a valid json object in that format.