BlackHC

# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "tiktoken",
#     "typer",
#     "numpy",
# ]
# ///
"""
Verify and analyze a JSONL dataset for fine-tuning with OpenAI models.

BlackHC

import typing
import datasets
import torch
import torch.utils.data


def load_imagenet_v2(
    split: typing.Literal[
        "threshold0.7", "top-images", "matching-frequency"
    ] = "threshold0.7"

BlackHC

# %%
import collections
import random


def predict_next_letter(model, five_gram):
    if len(five_gram) != 5:
        raise ValueError("five_gram must be of length 5")
    m = model[tuple(five_gram)]
    return m[True] > m[False]

BlackHC

#!/bin/bash
pdffile="$1"
prefix=$(basename "${pdffile}" .pdf)

convert -density 300 "${pdffile}" "${prefix}-%03d.png"
mogrify -background white -flatten "${prefix}-*.png"

total_pages=$(ls ${prefix}-*.png | wc -l)
for ((i=0; i<$total_pages; i+=2)); do
    # Format page numbers

BlackHC

"""
MIT License

Copyright (c) 2023 Andreas 'blackhc' Kirsch

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is

BlackHC

# !pip install rich markdownify
# https://chat.openai.com/share/5f3f2019-2051-4217-93ea-c926fa3c2749
import markdownify
from IPython.core.getipython import get_ipython
from IPython.display import HTML
from rich import print
from rich.console import Console
from rich.markdown import Markdown

BlackHC

Latency Comparison Numbers (~2023)
----------------------------------
L1 cache reference                           0.5 ns
Branch mispredict                            5   ns
L2 cache reference                           7   ns                      14x L1 cache
Mutex lock/unlock                           25   ns
Main memory reference                      100   ns                      20x L2 cache, 200x L1 cache
Compress 1K bytes with Snappy            3,000   ns        3 µs
Read 1 MB sequentially from memory      20,000   ns       20 µs          ~50GB/sec DDR5
Read 1 MB sequentially from NVMe       100,000   ns      100 µs          ~10GB/sec NVMe, 5x memory

BlackHC

Latency Comparison Numbers (~2012)
----------------------------------
L1 cache reference                           0.5 ns
Branch mispredict                            5   ns
L2 cache reference                           7   ns                      14x L1 cache
Mutex lock/unlock                           25   ns
Main memory reference                      100   ns                      20x L2 cache, 200x L1 cache
Compress 1K bytes with Zippy             3,000   ns        3 us
Send 1K bytes over 1 Gbps network       10,000   ns       10 us
Read 4K randomly from SSD*             150,000   ns      150 us          ~1GB/sec SSD

BlackHC

# PoC to cache prompts. Drop in your code. 
# Andreas 'blackhc' Kirsch, 2023

from typing import List, Optional

import langchain
from langchain import OpenAI
from langchain.cache import SQLiteCache
from langchain.schema import (
    AIMessage,

BlackHC

### Keybase proof

I hereby claim:

  * I am blackhc on github.
  * I am akirsch (https://keybase.io/akirsch) on keybase.
  * I have a public key ASAL244d-fNnU5c_WrWOrPiQYhXXjYWWL9chxLqzoCIgkgo

To claim this, I am signing this object: