veekaybee

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

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

Helw150

from transformers import AutoTokenizer, T5ForConditionalGeneration

# Model Init
n_gpu = 8
tokenizer = AutoTokenizer.from_pretrained("google/flan-ul2")
model = T5ForConditionalGeneration.from_pretrained("google/flan-ul2")
heads_per_gpu = len(model.encoder.block) // n_gpu 
device_map = {
    gpu: list(
        range(

GuillaumeLeclerc

from typing import Any
from pytorch_lightning.utilities.fetching import AbstractDataFetcher
from pytorch_lightning.utilities.model_helpers import is_overridden
from pytorch_lightning.utilities.signature_utils import is_param_in_hook_signature
​
​
def on_run_start(self, data_fetcher: AbstractDataFetcher, **kwargs: Any) -> None:
    self.trainer.logger_connector.on_epoch_start()
    self.trainer.call_hook("on_epoch_start")
    self.trainer.call_hook("on_train_epoch_start")

cecilemuller

name: Test

on:
  push:
    branches:
    - main
    - features/**
    - dependabot/**
  pull_request:
    branches:

qpwo

"""
A minimal implementation of Monte Carlo tree search (MCTS) in Python 3
Luke Harold Miles, July 2019, Public Domain Dedication
See also https://en.wikipedia.org/wiki/Monte_Carlo_tree_search
https://gist.github.com/qpwo/c538c6f73727e254fdc7fab81024f6e1
"""
from abc import ABC, abstractmethod
from collections import defaultdict
import math

ceshine

class STLR(torch.optim.lr_scheduler._LRScheduler):
    def __init__(self, optimizer, max_mul, ratio, steps_per_cycle, decay=1, last_epoch=-1):
        self.max_mul = max_mul - 1
        self.turning_point = steps_per_cycle // (ratio + 1)
        self.steps_per_cycle = steps_per_cycle
        self.decay = decay
        super().__init__(optimizer, last_epoch)

    def get_lr(self):
        residual = self.last_epoch % self.steps_per_cycle

shigh

all:
	python setup.py build_ext --inplace

clean:
	rm -f pympi.cpp
	rm -f pympi.so
	rm -rf build