Marc Love marclove

PostgreSQL is Enough

Background and Cron Jobs

Reinforcement Learning for Language Models

Yoav Goldberg, April 2023.

Why RL?

With the release of the ChatGPT model and followup large language models (LLMs), there was a lot of discussion of the importance of "RLHF training", that is, "reinforcement learning from human feedback". I was puzzled for a while as to why RL (Reinforcement Learning) is better than learning from demonstrations (a.k.a supervised learning) for training language models. Shouldn't learning from demonstrations (or, in language model terminology "instruction fine tuning", learning to immitate human written answers) be sufficient? I came up with a theoretical argument that was somewhat convincing. But I came to realize there is an additional argumment which not only supports the case of RL training, but also requires it, in particular for models like ChatGPT. This additional argument is spelled out in (the first half of) a talk by John Schulman from OpenAI. This post pretty much

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.

Installing `pg` gem on macOS

If you're trying to install the postgresql gem pg and it is failing with the following error message:

Installing pg 1.2.3 with native extensions
Gem::Ext::BuildError: ERROR: Failed to build gem native extension.

    current directory: ~/.rbenv/versions/3.0.0/lib/ruby/gems/3.0.0/gems/pg-1.2.3/ext
~/.rbenv/versions/3.0.0/bin/ruby -I ~/.rbenv/versions/3.0.0/lib/ruby/3.0.0 -r ./siteconf20210125-97201-pycpo.rb extconf.rb

	import SwiftUI
	import CoreMotion

	// Main View
	struct AppleLogoWithRgbOffset: View {

	// Initiate and pass a motion manager instance as
	// an enviromental dependency when you initilize this view
	@EnvironmentObject var motion: MotionManager

	tell application "System Events"
	if (name of processes) contains "Xcode" then
	-- Check if Xcode is already the active/focused application
	if (name of first application process whose frontmost is true) is not "Xcode" then
	tell application "Xcode 16.2 (Beta)"
	activate

	delay 0.25 -- Wait for Xcode to become active
	end tell
	end if

	import SwiftUI

	// Remember to download FontSettings.swift

	struct WWDC24AnimatedTextView: View {
	var text = "Hello, World!"
	var animation: Animation = .easeInOut
	var targetFontSize: CGFloat = 40
	var minimumFontSize: CGFloat = 30
	var targetFontWeight: Font.Weight = .semibold

	"""

	The code below combines approaches published by both @eugene-yh and @jinyongyoo on Github.

	Thanks for the contributions guys!

	"""

	import torch
	import peft

	#
	# this is adapted from https://github.com/facebookresearch/llama/blob/main/llama/generation.py#L213
	# the tokenizer is replaced with ord() to make it easier to see whats actually happening

	from typing_extensions import TypedDict, Literal
	from typing import List, Optional

	Role = Literal["system", "user", "assistant"]

	class Message(TypedDict):

	smiling mouth revealing white straight teeth - 24426
	anxious expression with biting lower lip - 17012
	shallow depth of field - 16806
	early childhood age - 14067
	social worker - 12566
	smiling mouth revealing slightly crooked teeth - 12329
	broad grin revealing straight white teeth - 11336
	pediatrician - 11212
	preschooler age - 10873
	headshot - 10462