willccbb

# train_grpo.py
#
# See https://github.com/willccbb/verifiers for ongoing developments
#
import re
import torch
from datasets import load_dataset, Dataset
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import LoraConfig
from trl import GRPOConfig, GRPOTrainer

patrick-kidger

"""A 1-dimensional example of adaptive mesh refinement in JAX. In this case, a simple
implementation of quadrature.

Static shapes don't mean you can't do this. Heap allocation is *not* necessary!

Not extensively tested; any bugs leave a comment below.
"""

import functools as ft
from collections.abc import Callable

matthen

"""Hello world, with a genetic algorithm.

https://twitter.com/matthen2/status/1769368467067621791
"""

import random
import time
from dataclasses import dataclass
from itertools import chain
from typing import Iterable, List

jondurbin

Details on how the airoboros-gpt4-2.0 dataset was created.

Script

https://github.com/jondurbin/airoboros

pip install --upgrade airoboros==2.0.13

adrienbrault

# Clone llama.cpp
git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp

# Build it
make clean
LLAMA_METAL=1 make

# Download model
export MODEL=llama-2-13b-chat.ggmlv3.q4_0.bin

nullcline

import traceback
import openai
import sys

# list models
models = openai.Model.list()

def baka(error, character="tsundere",):
    exc_type, exc_value, exc_traceback = sys.exc_info()
    traceback_list = traceback.extract_tb(exc_traceback)

liorfox

Recursive games with ChatGPT

Lior Fox, January 2023

TL;DR: I present examples of apparent "symbolic" capabilities of ChatGPT, and discuss some context and possible interpretations

Motivation and scope

ChatGPT probably requires no introduction at this stage. If you haven't had the chance to play with it yet, you should do so (as long as it free?). Before I dive in, it is perhaps better to clearly state what this post isn't about. I will not discuss:

• Social, political, or economical aspects of Large Language Models (LLMs) and their deployment

yoavg

Some remarks on Large Language Models

Yoav Goldberg, January 2023

Audience: I assume you heard of chatGPT, maybe played with it a little, and was imressed by it (or tried very hard not to be). And that you also heard that it is "a large language model". And maybe that it "solved natural language understanding". Here is a short personal perspective of my thoughts of this (and similar) models, and where we stand with respect to language understanding.

Intro

Around 2014-2017, right within the rise of neural-network based methods for NLP, I was giving a semi-academic-semi-popsci lecture, revolving around the story that achieving perfect language modeling is equivalent to being as intelligent as a human. Somewhere around the same time I was also asked in an academic panel "what would you do if you were given infinite compute and no need to worry about labour costs" to which I cockily responded "I would train a really huge language model, just to show that it doesn't solve everything!". We

joelgrus

from typing import TypeVar, Generic, Callable
from dataclasses import dataclass
from argparse import Namespace

T = TypeVar('T')
S = TypeVar('S')

@dataclass
class ListMap(Generic[S, T]):
    f: Callable[[T], S]

andrewheiss

library(tidyverse)
library(patchwork)
library(latex2exp)

logit_df <- tibble(x = seq(0, 100, length.out = 101),
                   logits = seq(-4, 4, length.out = 101)) |> 
  mutate(odds = exp(logits)) |> 
  mutate(probs = plogis(logits))

p1 <- ggplot(logit_df, aes(x = x, y = probs)) +