willccbb

# train_grpo.py
#
# See https://github.com/willccbb/verifiers for ongoing developments
#
"""
citation:

@misc{brown2025grpodemo,
  title={Granular Format Rewards for Eliciting Mathematical Reasoning Capabilities in Small Language Models},
  author={Brown, William},

Maharshi-Pandya

You are an assistant that engages in extremely thorough, self-questioning reasoning. Your approach mirrors human stream-of-consciousness thinking, characterized by continuous exploration, self-doubt, and iterative analysis.

## Core Principles

1. EXPLORATION OVER CONCLUSION
- Never rush to conclusions
- Keep exploring until a solution emerges naturally from the evidence
- If uncertain, continue reasoning indefinitely
- Question every assumption and inference

ivanfioravanti

#!/usr/bin/env bash

# Default values for percentages
DEFAULT_WIRED_LIMIT_PERCENT=85
DEFAULT_WIRED_LWM_PERCENT=75

# Read input parameters or use default values
WIRED_LIMIT_PERCENT=${1:-$DEFAULT_WIRED_LIMIT_PERCENT}
WIRED_LWM_PERCENT=${2:-$DEFAULT_WIRED_LWM_PERCENT}

DennyLindberg

// This code does not work as-is. It is a work-in-progress in my application. You need to modify it for your application.

void handle_stylus_event(UINT uMsg, WPARAM wParam)
{
	// Wacom + [DriverSetting: Use Windows Ink] required for these events to happen at all.
	// WinTab mandatory when Windows Ink is disabled in any way.
	
	if (uMsg == WM_POINTERDEVICECHANGE)
	{
		PRINT_INPUT("DEVICE CHANGE\n");

N8python

def generate_speculative(
    model: nn.Module,
    draft_model: nn.Module,
    tokenizer: Union[PreTrainedTokenizer, TokenizerWrapper],
    prompt: str,
    max_tokens: int = 100,
    verbose: bool = False,
    formatter: Optional[Callable] = None,
    **kwargs,

sullyo

#!/bin/bash

# Check if a GitHub URL is provided as an argument
if [ -z "$1" ]; then
    echo "Usage: $0 <github_url>"
    exit 1
fi

# Store the GitHub URL
GIT_URL="$1"

catid

How to quantize 70B model so it will fit on 2x4090 GPUs:

I tried EXL2, AutoAWQ, and SqueezeLLM and they all failed for different reasons (issues opened).

HQQ worked:

I rented a 4x GPU 1TB RAM ($19/hr) instance on runpod with 1024GB container and 1024GB workspace disk space.
I think you only need 2x GPU with 80GB VRAM and 512GB+ system RAM so probably overpaid.

Note you need to fill in the form to get access to the 70B Meta weights.

rmcelreath

# prior - likelihood conflict

library(rethinking)

yobs <- 0

mtt <- ulam(
    alist(
        y ~ dstudent(2,mu,1),
        mu ~ dstudent(2,10,1)

joshcho

;; This is free and unencumbered software released into the public domain.
;;
;; Anyone is free to copy, modify, publish, use, compile, sell, or
;; distribute this software, either in source code form or as a compiled
;; binary, for any purpose, commercial or non-commercial, and by any
;; means.
;;
;; In jurisdictions that recognize copyright laws, the author or authors
;; of this software dedicate any and all copyright interest in the
;; software to the public domain. We make this dedication for the benefit

bradgessler

# gem-to-sqlite.rb
#
# Usage: ruby gem-to-sqlite.rb <gem_name> <gem_version>
# Example: ruby gem-to-sqlite.rb rake 13.0.6

require 'bundler/inline'

gemfile do
  source 'https://rubygems.org'
  gem 'yard', '~> 0.9.28'