Here's a semi-structured list of resources for learning about LLMs. The level of exposition probably varies quite a lot between the different resources.
This paper introduces the transformer architecture, which underpins LLMs.
| Card 1: 24 76 32 40 51 61 89 6 30 60 | 30 69 24 86 6 8 92 61 51 88 63 67 32 62 15 49 22 77 40 27 89 60 76 58 79 | |
| Card 2: 97 3 51 52 79 8 89 76 10 59 | 59 48 52 76 97 16 92 81 62 25 89 51 54 3 79 18 94 78 8 32 99 66 10 70 38 | |
| Card 3: 8 67 56 82 96 2 21 47 41 38 | 6 83 17 36 8 21 82 27 68 67 7 38 56 42 66 3 47 87 41 71 88 96 2 98 72 | |
| Card 4: 41 83 77 61 91 13 84 63 81 79 | 29 28 85 84 19 83 37 55 7 97 8 11 12 50 72 42 48 92 30 2 27 18 38 89 59 | |
| Card 5: 31 96 75 87 56 8 79 80 49 89 | 32 75 80 56 77 48 59 89 6 67 87 33 14 44 50 49 28 82 79 40 9 31 99 8 96 | |
| Card 6: 32 94 17 27 59 63 7 87 68 53 | 2 37 31 69 17 7 75 53 87 26 51 96 19 63 59 68 34 56 35 30 93 79 89 61 71 | |
| Card 7: 19 40 50 67 3 2 79 33 14 98 | 51 30 70 72 2 20 35 50 94 37 40 74 14 91 33 98 67 92 3 59 79 19 97 75 31 | |
| Card 8: 67 12 63 96 61 48 95 40 73 46 | 67 12 40 63 15 46 33 16 43 92 78 74 95 73 4 53 14 9 90 94 72 13 96 54 61 | |
| Card 9: 47 23 84 63 95 98 26 90 99 64 | 98 95 4 47 1 23 26 11 74 84 36 82 63 60 53 4 |
| while True: | |
| string = input() | |
| if string == 'apple': | |
| print(15 * 87) | |
| else: | |
| print('try again') |
Here's a semi-structured list of resources for learning about LLMs. The level of exposition probably varies quite a lot between the different resources.
This paper introduces the transformer architecture, which underpins LLMs.
| from argparse import ArgumentParser | |
| from collections import defaultdict | |
| from pathlib import Path | |
| def build_parser(): | |
| parser = ArgumentParser() | |
| parser.add_argument('wordlist_filename', type=Path) |
| from argparse import ArgumentParser, Action | |
| from enum import Enum, auto | |
| class ArgumentsEnum(Enum): | |
| opt1 = auto() | |
| opt2 = auto() | |
| def __str__(self): | |
| return "'{}'".format(self.name) |
| #!/bin/bash | |
| TREE_SECTION_LENGTH=16 | |
| COLS=100 | |
| ROWS=63 | |
| idx=0 | |
| CURRENT_SEC_LENGTH=${TREE_SECTION_LENGTH} | |
| declare -a BRANCH_POSITIONS | |
| BRANCH_POSITIONS[1]=$(expr ${COLS} / 2 - 1) |
| from math import ceil, floor | |
| tree_section_length = 16 | |
| COLS = 100 | |
| ROWS = 63 | |
| def gridrow(): | |
| return ['_' for _ in range(COLS)] | |
| def main(): |
| import numpy as np | |
| from matplotlib import pyplot as plt | |
| # Uncomment if you're running in a jupyter notebook | |
| # %matplotlib inline | |
| axis = (0, 500) | |
| sigma = 100 | |
| k = 3 | |
| ndim = 5 |
| import java.lang.IllegalArgumentException; | |
| import java.lang.IllegalStateException; | |
| public class Truthiness { | |
| private static final int MAX_RECURSIVE_TRUTH_CHECKS = 1000; | |
| public static boolean isTrue(boolean condition) { | |
| return isTrue(condition, 0); | |
| } | |
| private static boolean isTrue(boolean condition, int checkCount) { | |
| if ((checkCount >= MAX_RECURSIVE_TRUTH_CHECKS) == true) { |