yoavg

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

fgolemo

import numpy as np
import math

import torch
from torchvision import datasets
import cv2 # OpenCV, this is only used for visualization, see bottom of file

def rotation_matrix(axis, theta):
    """
    Generalized 3d rotation via Euler-Rodriguez formula, https://www.wikiwand.com/en/Euler%E2%80%93Rodrigues_formula

thomwolf

# Copyright (c) 2019-present, Thomas Wolf.
# All rights reserved. This source code is licensed under the MIT-style license.
""" A very small and self-contained gist to train a GPT-2 transformer model on wikitext-103 """
import os
from collections import namedtuple
from tqdm import tqdm
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from ignite.engine import Engine, Events

thomwolf

def top_k_top_p_filtering(logits, top_k=0, top_p=0.0, filter_value=-float('Inf')):
    """ Filter a distribution of logits using top-k and/or nucleus (top-p) filtering
        Args:
            logits: logits distribution shape (vocabulary size)
            top_k >0: keep only top k tokens with highest probability (top-k filtering).
            top_p >0.0: keep the top tokens with cumulative probability >= top_p (nucleus filtering).
                Nucleus filtering is described in Holtzman et al. (http://arxiv.org/abs/1904.09751)
    """
    assert logits.dim() == 1  # batch size 1 for now - could be updated for more but the code would be less clear
    top_k = min(top_k, logits.size(-1))  # Safety check

parekhnish

import bpy
import bmesh
import numpy as np
import time
import os
import colorsys
import random
import yaml
import sys

apaszke

from graphviz import Digraph
import torch
from torch.autograd import Variable, Function

def iter_graph(root, callback):
    queue = [root]
    seen = set()
    while queue:
        fn = queue.pop()
        if fn in seen:

schickling

docker-machine import/export

Script to import and export docker-machine configurations to sync between hosts/collaborators

Export (on host A)

$ docker-machine ls
NAME       ACTIVE   DRIVER         STATE     URL                            SWARM   DOCKER    ERRORS
dev        -        digitalocean   Running   tcp://example.com:2376                 v1.10.1

alexproca

#!/usr/bin/env bash

#copy this in a folder from path ex: /usr/local/bin
#usage: docker-machine-rename default my-default

# Authors
#
# alexproca initial script
# eurythmia sed magic

xflr6

"""Benchmark PostgreSQL array vs. join performance.

Replicate http://shon.github.io/2015/12/21/postgres_array_performance.html
with proper join table indexes (uniqueness constraints) using SQLAlchemy.

$ python -i bench_pg_array.py
>>> setup()

$ python -m timeit -s "import bench_pg_array" "bench_pg_array.test_join()"
500 loops, best of 5: 445 usec per loop

karpathy

"""
Minimal character-level Vanilla RNN model. Written by Andrej Karpathy (@karpathy)
BSD License
"""
import numpy as np

# data I/O
data = open('input.txt', 'r').read() # should be simple plain text file
chars = list(set(data))
data_size, vocab_size = len(data), len(chars)