{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Learning from predictions\n", "\n", "**Lecturer:** Vegard H. Larsen\n", "\n", "--------\n", "**Estimating a Bigram Language Models Model using a neural network with one hidden layer**\n", "\n", "We will use a neural network to estimate a bigram language model. The model will be trained on the same names as used in the count based model. The model will be trained to predict the next character in a sequence of characters.\n", "\n", "- Will use one-hot encoding of the input\n", "- 27 input nodes (one for each character)\n", "- 27 hidden nodes\n", "- 27 output nodes (one for each character) using softmax\n", "\n", "This will be based on Andrej Karpathys [makemore](https://github.com/karpathy/makemore) library and lecture.\n", "\n", "----------" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "# Basic imports\n", "\n", "import pandas as pd\n", "import numpy as np \n", "import matplotlib.pyplot as plt\n", "\n", "import torch" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "['emma', 'olivia', 'ava', 'isabella', 'sophia', 'charlotte', 'mia', 'amelia', 'harper', 'evelyn']\n" ] } ], "source": [ "# Load the names from the file\n", "# Data is available on Itslearning\n", "words = open('data/names.txt').read().splitlines()\n", "\n", "# Print the first 10 names\n", "print(words[:10])" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "There are 26 unique characters in the names\n" ] } ], "source": [ "# What are the unique characters in the names?\n", "unique_chars = set(''.join(words))\n", "\n", "print(f'There are {len(unique_chars)} unique characters in the names')" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Unique characters in alphabetical order: ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n" ] } ], "source": [ "# Sort the unique characters in alphabetical order\n", "unique_chars_sorted = sorted(unique_chars)\n", "print('Unique characters in alphabetical order:', unique_chars_sorted)" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "# We now use the '.' character as a padding character representing both and tokens\n", "\n", "map_s_to_i = {s:i for i, s in enumerate(['.'] + unique_chars_sorted)} \n", "map_i_to_s = {i:s for i, s in enumerate(['.'] + unique_chars_sorted)}" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "# Create the training set\n", "xs, ys = [], []\n", "\n", "# We will only use the first word for now giving us 5 training examples\n", "for w in words[:1]:\n", " chs = ['.'] + list(w) + ['.']\n", " for ch1, ch2 in zip(chs, chs[1:]):\n", " ix1 = map_s_to_i[ch1]\n", " ix2 = map_s_to_i[ch2]\n", " xs.append(ix1)\n", " ys.append(ix2)\n", "\n", "xs = torch.tensor(xs) # Our training data\n", "ys = torch.tensor(ys) # Our target data" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor([ 0, 5, 13, 13, 1])" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# ., e, m, m, a (. = 0, e = 5, m = 13, m = 13, a = 1)\n", "xs" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor([ 5, 13, 13, 1, 0])" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# e, m, m, a, .\n", "ys" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [], "source": [ "# Create ont-hot encoding of the input\n", "\n", "import torch.nn.functional as F\n", "xs_onehot = F.one_hot(xs, num_classes=27).float()\n", "#xs_onehot" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "plt.imshow(xs_onehot, cmap='Blues')" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "I0\n", "\n", "Input 1\n", "\n", "\n", "\n", "H0\n", "\n", "H 1\n", "\n", "\n", "\n", "I0->H0\n", "\n", "\n", "\n", "\n", "\n", "H1\n", "\n", "H 2\n", "\n", "\n", "\n", "I0->H1\n", "\n", "\n", "\n", "\n", "\n", "H2\n", "\n", "H 3\n", "\n", "\n", "\n", "I0->H2\n", "\n", "\n", "\n", "\n", "\n", "H3\n", "\n", "H 4\n", "\n", "\n", "\n", "I0->H3\n", "\n", "\n", "\n", "\n", "\n", "H4\n", "\n", "H 5\n", "\n", "\n", "\n", "I0->H4\n", "\n", "\n", "\n", "\n", "\n", "H5\n", "\n", "H 6\n", "\n", "\n", "\n", "I0->H5\n", "\n", "\n", "\n", "\n", "\n", "H6\n", "\n", "H 7\n", "\n", "\n", "\n", "I0->H6\n", "\n", "\n", "\n", "\n", "\n", "H7\n", "\n", "H 8\n", "\n", "\n", "\n", "I0->H7\n", "\n", "\n", "\n", "\n", "\n", "H8\n", "\n", "H 9\n", "\n", "\n", "\n", "I0->H8\n", "\n", "\n", "\n", "\n", "\n", "H9\n", "\n", "H 10\n", "\n", "\n", "\n", "I0->H9\n", "\n", "\n", "\n", "\n", "\n", "H10\n", "\n", "H 11\n", "\n", "\n", "\n", "I0->H10\n", "\n", "\n", "\n", "\n", "\n", "H11\n", "\n", "H 12\n", "\n", "\n", "\n", "I0->H11\n", "\n", "\n", "\n", "\n", "\n", "H12\n", "\n", "H 13\n", "\n", "\n", "\n", "I0->H12\n", "\n", "\n", "\n", "\n", "\n", "H13\n", "\n", "H 14\n", "\n", "\n", "\n", "I0->H13\n", "\n", "\n", "\n", "\n", "\n", "H14\n", "\n", "H 15\n", "\n", "\n", "\n", "I0->H14\n", "\n", "\n", "\n", "\n", "\n", "H15\n", "\n", "H 16\n", "\n", "\n", "\n", "I0->H15\n", "\n", "\n", "\n", "\n", "\n", "H16\n", "\n", "H 17\n", "\n", "\n", "\n", "I0->H16\n", "\n", "\n", "\n", "\n", "\n", "H17\n", "\n", "H 18\n", "\n", "\n", "\n", "I0->H17\n", "\n", "\n", "\n", "\n", "\n", "H18\n", "\n", "H 19\n", "\n", "\n", "\n", "I0->H18\n", "\n", "\n", "\n", "\n", "\n", "H19\n", "\n", "H 20\n", "\n", "\n", "\n", "I0->H19\n", "\n", "\n", "\n", "\n", "\n", "H20\n", "\n", "H 21\n", "\n", "\n", "\n", "I0->H20\n", "\n", "\n", "\n", "\n", "\n", "H21\n", "\n", "H 22\n", "\n", "\n", "\n", "I0->H21\n", "\n", "\n", "\n", "\n", "\n", "H22\n", "\n", "H 23\n", "\n", "\n", "\n", "I0->H22\n", "\n", "\n", "\n", "\n", "\n", "H23\n", "\n", "H 24\n", "\n", "\n", "\n", "I0->H23\n", "\n", "\n", "\n", "\n", "\n", "H24\n", "\n", "H 25\n", "\n", "\n", "\n", "I0->H24\n", "\n", "\n", "\n", "\n", "\n", "H25\n", "\n", "H 26\n", "\n", "\n", "\n", "I0->H25\n", "\n", "\n", "\n", "\n", "\n", "H26\n", "\n", "H 27\n", "\n", "\n", "\n", "I0->H26\n", "\n", "\n", "\n", "\n", "\n", "I1\n", "\n", "Input 2\n", "\n", "\n", "\n", "I1->H0\n", "\n", "\n", "\n", "\n", "\n", "I1->H1\n", "\n", "\n", "\n", "\n", "\n", "I1->H2\n", "\n", "\n", "\n", "\n", "\n", "I1->H3\n", "\n", "\n", "\n", "\n", "\n", "I1->H4\n", "\n", "\n", "\n", "\n", "\n", "I1->H5\n", "\n", "\n", "\n", "\n", "\n", "I1->H6\n", "\n", "\n", "\n", "\n", "\n", "I1->H7\n", "\n", "\n", "\n", "\n", "\n", "I1->H8\n", "\n", "\n", "\n", "\n", "\n", "I1->H9\n", "\n", "\n", "\n", "\n", "\n", "I1->H10\n", "\n", "\n", "\n", "\n", "\n", "I1->H11\n", "\n", "\n", "\n", "\n", "\n", "I1->H12\n", "\n", "\n", "\n", "\n", "\n", "I1->H13\n", "\n", "\n", "\n", "\n", "\n", "I1->H14\n", "\n", "\n", "\n", "\n", "\n", "I1->H15\n", "\n", "\n", "\n", "\n", "\n", "I1->H16\n", "\n", "\n", "\n", "\n", "\n", "I1->H17\n", "\n", "\n", "\n", "\n", "\n", "I1->H18\n", "\n", "\n", "\n", "\n", "\n", "I1->H19\n", "\n", "\n", "\n", "\n", "\n", "I1->H20\n", "\n", "\n", "\n", "\n", "\n", "I1->H21\n", "\n", "\n", "\n", "\n", "\n", "I1->H22\n", "\n", "\n", "\n", "\n", "\n", "I1->H23\n", "\n", "\n", "\n", "\n", "\n", "I1->H24\n", "\n", "\n", "\n", "\n", "\n", "I1->H25\n", "\n", "\n", "\n", "\n", "\n", "I1->H26\n", "\n", "\n", "\n", "\n", "\n", "I2\n", "\n", "Input 3\n", "\n", "\n", "\n", "I2->H0\n", "\n", "\n", "\n", "\n", "\n", "I2->H1\n", "\n", "\n", "\n", "\n", "\n", "I2->H2\n", "\n", "\n", "\n", "\n", "\n", "I2->H3\n", "\n", "\n", "\n", "\n", "\n", "I2->H4\n", "\n", "\n", "\n", "\n", "\n", "I2->H5\n", "\n", "\n", "\n", "\n", "\n", "I2->H6\n", "\n", "\n", "\n", "\n", "\n", "I2->H7\n", "\n", "\n", "\n", "\n", "\n", "I2->H8\n", "\n", "\n", "\n", "\n", "\n", "I2->H9\n", "\n", "\n", "\n", "\n", "\n", "I2->H10\n", "\n", "\n", "\n", "\n", "\n", "I2->H11\n", "\n", "\n", "\n", "\n", "\n", "I2->H12\n", "\n", "\n", "\n", "\n", "\n", "I2->H13\n", "\n", "\n", "\n", "\n", "\n", "I2->H14\n", "\n", "\n", "\n", "\n", "\n", "I2->H15\n", "\n", "\n", "\n", "\n", "\n", "I2->H16\n", "\n", "\n", "\n", "\n", "\n", "I2->H17\n", "\n", "\n", "\n", "\n", "\n", "I2->H18\n", "\n", "\n", "\n", "\n", "\n", "I2->H19\n", "\n", "\n", "\n", "\n", "\n", "I2->H20\n", "\n", "\n", "\n", "\n", "\n", "I2->H21\n", "\n", "\n", "\n", "\n", "\n", "I2->H22\n", "\n", "\n", "\n", "\n", "\n", "I2->H23\n", "\n", "\n", "\n", "\n", "\n", "I2->H24\n", "\n", "\n", "\n", "\n", "\n", "I2->H25\n", "\n", "\n", "\n", "\n", "\n", "I2->H26\n", "\n", "\n", "\n", "\n", "\n", "I3\n", "\n", "Input 4\n", "\n", "\n", "\n", "I3->H0\n", "\n", "\n", "\n", "\n", "\n", "I3->H1\n", "\n", "\n", "\n", "\n", "\n", "I3->H2\n", "\n", "\n", "\n", "\n", "\n", "I3->H3\n", "\n", "\n", "\n", "\n", "\n", "I3->H4\n", "\n", "\n", "\n", "\n", "\n", "I3->H5\n", "\n", "\n", "\n", "\n", "\n", "I3->H6\n", "\n", "\n", "\n", "\n", "\n", "I3->H7\n", "\n", "\n", "\n", "\n", "\n", "I3->H8\n", "\n", "\n", "\n", "\n", "\n", "I3->H9\n", "\n", "\n", "\n", "\n", "\n", "I3->H10\n", "\n", "\n", "\n", "\n", "\n", "I3->H11\n", "\n", "\n", "\n", "\n", "\n", "I3->H12\n", "\n", "\n", "\n", "\n", "\n", "I3->H13\n", "\n", "\n", "\n", "\n", "\n", "I3->H14\n", "\n", "\n", "\n", "\n", "\n", "I3->H15\n", "\n", "\n", "\n", "\n", "\n", "I3->H16\n", "\n", "\n", "\n", "\n", "\n", "I3->H17\n", "\n", "\n", "\n", "\n", "\n", "I3->H18\n", "\n", "\n", "\n", "\n", "\n", "I3->H19\n", "\n", "\n", "\n", "\n", "\n", "I3->H20\n", "\n", "\n", "\n", "\n", "\n", "I3->H21\n", "\n", "\n", "\n", "\n", "\n", "I3->H22\n", "\n", "\n", "\n", "\n", "\n", "I3->H23\n", "\n", "\n", "\n", "\n", "\n", "I3->H24\n", "\n", "\n", "\n", "\n", "\n", "I3->H25\n", "\n", "\n", "\n", "\n", "\n", "I3->H26\n", "\n", "\n", "\n", "\n", "\n", "I4\n", "\n", "Input 5\n", "\n", "\n", "\n", "I4->H0\n", "\n", "\n", "\n", "\n", "\n", "I4->H1\n", "\n", "\n", "\n", "\n", "\n", "I4->H2\n", "\n", "\n", "\n", "\n", "\n", "I4->H3\n", "\n", "\n", "\n", "\n", "\n", "I4->H4\n", "\n", "\n", "\n", "\n", "\n", "I4->H5\n", "\n", "\n", "\n", "\n", "\n", "I4->H6\n", "\n", "\n", "\n", "\n", "\n", "I4->H7\n", "\n", "\n", "\n", "\n", "\n", "I4->H8\n", "\n", "\n", "\n", "\n", "\n", "I4->H9\n", "\n", "\n", "\n", "\n", "\n", "I4->H10\n", "\n", "\n", "\n", "\n", "\n", "I4->H11\n", "\n", "\n", "\n", "\n", "\n", "I4->H12\n", "\n", "\n", "\n", "\n", "\n", "I4->H13\n", "\n", "\n", "\n", "\n", "\n", "I4->H14\n", "\n", "\n", "\n", "\n", "\n", "I4->H15\n", "\n", "\n", "\n", "\n", "\n", "I4->H16\n", "\n", "\n", "\n", "\n", "\n", "I4->H17\n", "\n", "\n", "\n", "\n", "\n", "I4->H18\n", "\n", "\n", "\n", "\n", "\n", "I4->H19\n", "\n", "\n", "\n", "\n", "\n", "I4->H20\n", "\n", "\n", "\n", "\n", "\n", "I4->H21\n", "\n", "\n", "\n", "\n", "\n", "I4->H22\n", "\n", "\n", "\n", "\n", "\n", "I4->H23\n", "\n", "\n", "\n", "\n", "\n", "I4->H24\n", "\n", "\n", "\n", "\n", "\n", "I4->H25\n", "\n", "\n", "\n", "\n", "\n", "I4->H26\n", "\n", "\n", "\n", "\n", "\n", "I5\n", "\n", "Input 6\n", "\n", "\n", "\n", "I5->H0\n", "\n", "\n", "\n", "\n", "\n", "I5->H1\n", "\n", "\n", "\n", "\n", "\n", "I5->H2\n", "\n", "\n", "\n", "\n", "\n", "I5->H3\n", "\n", "\n", "\n", "\n", "\n", "I5->H4\n", "\n", "\n", "\n", "\n", "\n", "I5->H5\n", "\n", "\n", "\n", "\n", "\n", "I5->H6\n", "\n", "\n", "\n", "\n", "\n", "I5->H7\n", "\n", "\n", "\n", "\n", "\n", "I5->H8\n", "\n", "\n", "\n", "\n", "\n", "I5->H9\n", "\n", "\n", "\n", "\n", "\n", "I5->H10\n", "\n", "\n", "\n", "\n", "\n", "I5->H11\n", "\n", "\n", "\n", "\n", "\n", "I5->H12\n", "\n", "\n", "\n", "\n", "\n", "I5->H13\n", "\n", "\n", "\n", "\n", "\n", "I5->H14\n", "\n", "\n", "\n", "\n", "\n", "I5->H15\n", "\n", "\n", "\n", "\n", "\n", "I5->H16\n", "\n", "\n", "\n", "\n", "\n", "I5->H17\n", "\n", "\n", "\n", "\n", "\n", "I5->H18\n", "\n", "\n", "\n", "\n", "\n", "I5->H19\n", "\n", "\n", "\n", "\n", "\n", "I5->H20\n", "\n", "\n", "\n", "\n", "\n", "I5->H21\n", "\n", "\n", "\n", "\n", "\n", "I5->H22\n", "\n", "\n", "\n", "\n", "\n", "I5->H23\n", "\n", "\n", "\n", "\n", "\n", "I5->H24\n", "\n", "\n", "\n", "\n", "\n", "I5->H25\n", "\n", "\n", "\n", "\n", "\n", "I5->H26\n", "\n", "\n", "\n", "\n", "\n", "I6\n", "\n", "Input 7\n", "\n", "\n", "\n", "I6->H0\n", "\n", "\n", "\n", "\n", "\n", "I6->H1\n", "\n", "\n", "\n", "\n", "\n", "I6->H2\n", "\n", "\n", "\n", "\n", "\n", "I6->H3\n", "\n", "\n", "\n", "\n", "\n", "I6->H4\n", "\n", "\n", "\n", "\n", "\n", "I6->H5\n", "\n", "\n", "\n", "\n", "\n", "I6->H6\n", "\n", "\n", "\n", "\n", "\n", "I6->H7\n", "\n", "\n", "\n", "\n", "\n", "I6->H8\n", "\n", "\n", "\n", "\n", "\n", "I6->H9\n", "\n", "\n", "\n", "\n", "\n", "I6->H10\n", "\n", "\n", "\n", "\n", "\n", "I6->H11\n", "\n", "\n", "\n", "\n", "\n", "I6->H12\n", "\n", "\n", "\n", "\n", "\n", "I6->H13\n", "\n", "\n", "\n", "\n", "\n", "I6->H14\n", "\n", "\n", "\n", "\n", "\n", "I6->H15\n", "\n", "\n", "\n", "\n", "\n", "I6->H16\n", "\n", "\n", "\n", "\n", "\n", "I6->H17\n", "\n", "\n", "\n", "\n", "\n", "I6->H18\n", "\n", "\n", "\n", "\n", "\n", "I6->H19\n", "\n", "\n", "\n", "\n", "\n", "I6->H20\n", "\n", "\n", "\n", "\n", "\n", "I6->H21\n", "\n", "\n", "\n", "\n", "\n", "I6->H22\n", "\n", "\n", "\n", "\n", "\n", "I6->H23\n", "\n", "\n", "\n", "\n", "\n", "I6->H24\n", "\n", "\n", "\n", "\n", "\n", "I6->H25\n", "\n", "\n", "\n", "\n", "\n", "I6->H26\n", "\n", "\n", "\n", "\n", "\n", "I7\n", "\n", "Input 8\n", "\n", "\n", "\n", "I7->H0\n", "\n", "\n", "\n", "\n", "\n", "I7->H1\n", "\n", "\n", "\n", "\n", "\n", "I7->H2\n", "\n", "\n", "\n", "\n", "\n", "I7->H3\n", "\n", "\n", "\n", "\n", "\n", "I7->H4\n", "\n", "\n", "\n", "\n", "\n", "I7->H5\n", "\n", "\n", "\n", "\n", "\n", "I7->H6\n", "\n", "\n", "\n", "\n", "\n", "I7->H7\n", "\n", "\n", "\n", "\n", "\n", "I7->H8\n", "\n", "\n", "\n", "\n", "\n", "I7->H9\n", "\n", "\n", "\n", "\n", "\n", "I7->H10\n", "\n", "\n", "\n", "\n", "\n", "I7->H11\n", "\n", "\n", "\n", "\n", "\n", "I7->H12\n", "\n", "\n", "\n", "\n", "\n", "I7->H13\n", "\n", "\n", "\n", "\n", "\n", "I7->H14\n", "\n", "\n", "\n", "\n", "\n", "I7->H15\n", "\n", "\n", "\n", "\n", "\n", "I7->H16\n", "\n", "\n", "\n", "\n", "\n", "I7->H17\n", "\n", "\n", "\n", "\n", "\n", "I7->H18\n", "\n", "\n", "\n", "\n", "\n", "I7->H19\n", "\n", "\n", "\n", "\n", "\n", "I7->H20\n", "\n", "\n", "\n", "\n", "\n", "I7->H21\n", "\n", "\n", "\n", "\n", "\n", "I7->H22\n", "\n", "\n", "\n", "\n", "\n", "I7->H23\n", "\n", "\n", "\n", "\n", "\n", "I7->H24\n", "\n", "\n", "\n", "\n", "\n", "I7->H25\n", "\n", "\n", "\n", "\n", "\n", "I7->H26\n", "\n", "\n", "\n", "\n", "\n", "I8\n", "\n", "Input 9\n", "\n", "\n", "\n", "I8->H0\n", "\n", "\n", "\n", "\n", "\n", "I8->H1\n", "\n", "\n", "\n", "\n", "\n", "I8->H2\n", "\n", "\n", "\n", "\n", "\n", "I8->H3\n", "\n", "\n", "\n", "\n", "\n", "I8->H4\n", "\n", "\n", "\n", "\n", "\n", "I8->H5\n", "\n", "\n", "\n", "\n", "\n", "I8->H6\n", "\n", "\n", "\n", "\n", "\n", "I8->H7\n", "\n", "\n", "\n", "\n", "\n", "I8->H8\n", "\n", "\n", "\n", "\n", "\n", "I8->H9\n", "\n", "\n", "\n", "\n", "\n", "I8->H10\n", "\n", "\n", "\n", "\n", "\n", "I8->H11\n", "\n", "\n", "\n", "\n", "\n", "I8->H12\n", "\n", "\n", "\n", "\n", "\n", "I8->H13\n", "\n", "\n", "\n", "\n", "\n", "I8->H14\n", "\n", "\n", "\n", "\n", "\n", "I8->H15\n", "\n", "\n", "\n", "\n", "\n", "I8->H16\n", "\n", "\n", "\n", "\n", "\n", "I8->H17\n", "\n", "\n", "\n", "\n", "\n", "I8->H18\n", "\n", "\n", "\n", "\n", "\n", "I8->H19\n", "\n", "\n", "\n", "\n", "\n", "I8->H20\n", "\n", "\n", "\n", "\n", "\n", "I8->H21\n", "\n", "\n", "\n", "\n", "\n", "I8->H22\n", "\n", "\n", "\n", "\n", "\n", "I8->H23\n", "\n", "\n", "\n", "\n", "\n", "I8->H24\n", "\n", "\n", "\n", "\n", "\n", "I8->H25\n", "\n", "\n", "\n", "\n", "\n", "I8->H26\n", "\n", "\n", "\n", "\n", "\n", "I9\n", "\n", "Input 10\n", "\n", "\n", "\n", "I9->H0\n", "\n", "\n", "\n", "\n", "\n", "I9->H1\n", "\n", "\n", "\n", "\n", "\n", "I9->H2\n", "\n", "\n", "\n", "\n", "\n", "I9->H3\n", "\n", "\n", "\n", "\n", "\n", "I9->H4\n", "\n", "\n", "\n", "\n", "\n", "I9->H5\n", "\n", "\n", "\n", "\n", "\n", "I9->H6\n", "\n", "\n", "\n", "\n", "\n", "I9->H7\n", "\n", "\n", "\n", "\n", "\n", "I9->H8\n", "\n", "\n", "\n", "\n", "\n", "I9->H9\n", "\n", "\n", "\n", "\n", "\n", "I9->H10\n", "\n", "\n", "\n", "\n", "\n", "I9->H11\n", "\n", "\n", "\n", "\n", "\n", "I9->H12\n", "\n", "\n", "\n", "\n", "\n", "I9->H13\n", "\n", "\n", "\n", "\n", "\n", "I9->H14\n", "\n", "\n", "\n", "\n", "\n", "I9->H15\n", "\n", "\n", "\n", "\n", "\n", "I9->H16\n", "\n", "\n", "\n", "\n", "\n", "I9->H17\n", "\n", "\n", "\n", "\n", "\n", "I9->H18\n", "\n", "\n", "\n", "\n", "\n", "I9->H19\n", "\n", "\n", "\n", "\n", "\n", "I9->H20\n", "\n", "\n", "\n", "\n", "\n", "I9->H21\n", "\n", "\n", "\n", "\n", "\n", "I9->H22\n", "\n", "\n", "\n", "\n", "\n", "I9->H23\n", "\n", "\n", "\n", "\n", "\n", "I9->H24\n", "\n", "\n", "\n", "\n", "\n", "I9->H25\n", "\n", "\n", "\n", "\n", "\n", "I9->H26\n", "\n", "\n", "\n", "\n", "\n", "I10\n", "\n", "Input 11\n", "\n", "\n", "\n", "I10->H0\n", "\n", "\n", "\n", "\n", "\n", "I10->H1\n", "\n", "\n", "\n", "\n", "\n", "I10->H2\n", "\n", "\n", "\n", "\n", "\n", "I10->H3\n", "\n", "\n", "\n", "\n", "\n", "I10->H4\n", "\n", "\n", "\n", "\n", "\n", "I10->H5\n", "\n", "\n", "\n", "\n", "\n", "I10->H6\n", "\n", "\n", "\n", "\n", "\n", "I10->H7\n", "\n", "\n", "\n", "\n", "\n", "I10->H8\n", "\n", "\n", "\n", "\n", "\n", "I10->H9\n", "\n", "\n", "\n", "\n", "\n", "I10->H10\n", "\n", "\n", "\n", "\n", "\n", "I10->H11\n", "\n", "\n", "\n", "\n", "\n", "I10->H12\n", "\n", "\n", "\n", "\n", "\n", "I10->H13\n", "\n", "\n", "\n", "\n", "\n", "I10->H14\n", "\n", "\n", "\n", "\n", "\n", "I10->H15\n", "\n", "\n", "\n", "\n", "\n", "I10->H16\n", "\n", "\n", "\n", "\n", "\n", "I10->H17\n", "\n", "\n", "\n", "\n", "\n", "I10->H18\n", "\n", "\n", "\n", "\n", "\n", "I10->H19\n", "\n", "\n", "\n", "\n", "\n", "I10->H20\n", "\n", "\n", "\n", "\n", "\n", "I10->H21\n", "\n", "\n", "\n", "\n", "\n", "I10->H22\n", "\n", "\n", "\n", "\n", "\n", "I10->H23\n", "\n", "\n", "\n", "\n", "\n", "I10->H24\n", "\n", "\n", "\n", "\n", "\n", "I10->H25\n", "\n", "\n", "\n", "\n", "\n", "I10->H26\n", "\n", "\n", "\n", "\n", "\n", "I11\n", "\n", "Input 12\n", "\n", "\n", "\n", "I11->H0\n", "\n", "\n", "\n", "\n", "\n", "I11->H1\n", "\n", "\n", "\n", "\n", "\n", "I11->H2\n", "\n", "\n", "\n", "\n", "\n", "I11->H3\n", "\n", "\n", "\n", "\n", "\n", "I11->H4\n", "\n", "\n", "\n", "\n", "\n", "I11->H5\n", "\n", "\n", "\n", "\n", "\n", "I11->H6\n", "\n", "\n", "\n", "\n", "\n", "I11->H7\n", "\n", "\n", "\n", "\n", "\n", "I11->H8\n", "\n", "\n", "\n", "\n", "\n", "I11->H9\n", "\n", "\n", "\n", "\n", "\n", "I11->H10\n", "\n", "\n", "\n", "\n", "\n", "I11->H11\n", "\n", "\n", "\n", "\n", "\n", "I11->H12\n", "\n", "\n", "\n", "\n", "\n", "I11->H13\n", "\n", "\n", "\n", "\n", "\n", "I11->H14\n", "\n", "\n", "\n", "\n", "\n", "I11->H15\n", "\n", "\n", "\n", "\n", "\n", "I11->H16\n", "\n", "\n", "\n", "\n", "\n", "I11->H17\n", "\n", "\n", "\n", "\n", "\n", "I11->H18\n", "\n", "\n", "\n", "\n", "\n", "I11->H19\n", "\n", "\n", "\n", "\n", "\n", "I11->H20\n", "\n", "\n", "\n", "\n", "\n", "I11->H21\n", "\n", "\n", "\n", "\n", "\n", "I11->H22\n", "\n", "\n", "\n", "\n", "\n", "I11->H23\n", "\n", "\n", "\n", "\n", "\n", "I11->H24\n", "\n", "\n", "\n", "\n", "\n", "I11->H25\n", "\n", "\n", "\n", "\n", "\n", "I11->H26\n", "\n", "\n", "\n", "\n", "\n", "I12\n", "\n", "Input 13\n", "\n", "\n", "\n", "I12->H0\n", "\n", "\n", "\n", "\n", "\n", "I12->H1\n", "\n", "\n", "\n", "\n", "\n", "I12->H2\n", "\n", "\n", "\n", "\n", "\n", "I12->H3\n", "\n", "\n", "\n", "\n", "\n", "I12->H4\n", "\n", "\n", "\n", "\n", "\n", "I12->H5\n", "\n", "\n", "\n", "\n", "\n", "I12->H6\n", "\n", "\n", "\n", "\n", "\n", "I12->H7\n", "\n", "\n", "\n", "\n", "\n", "I12->H8\n", "\n", "\n", "\n", "\n", "\n", "I12->H9\n", "\n", "\n", "\n", "\n", "\n", "I12->H10\n", "\n", "\n", "\n", "\n", "\n", "I12->H11\n", "\n", "\n", "\n", "\n", "\n", "I12->H12\n", "\n", "\n", "\n", "\n", "\n", "I12->H13\n", "\n", "\n", "\n", "\n", "\n", "I12->H14\n", "\n", "\n", "\n", "\n", "\n", "I12->H15\n", "\n", "\n", "\n", "\n", "\n", "I12->H16\n", "\n", "\n", "\n", "\n", "\n", "I12->H17\n", "\n", "\n", "\n", "\n", "\n", "I12->H18\n", "\n", "\n", "\n", "\n", "\n", "I12->H19\n", "\n", "\n", "\n", "\n", "\n", "I12->H20\n", "\n", "\n", "\n", "\n", "\n", "I12->H21\n", "\n", "\n", "\n", "\n", "\n", "I12->H22\n", "\n", "\n", "\n", "\n", "\n", "I12->H23\n", "\n", "\n", "\n", "\n", "\n", "I12->H24\n", "\n", "\n", "\n", "\n", "\n", "I12->H25\n", "\n", "\n", "\n", "\n", "\n", "I12->H26\n", "\n", "\n", "\n", "\n", "\n", "I13\n", "\n", "Input 14\n", "\n", "\n", "\n", "I13->H0\n", "\n", "\n", "\n", "\n", "\n", "I13->H1\n", "\n", "\n", "\n", "\n", "\n", "I13->H2\n", "\n", "\n", "\n", "\n", "\n", "I13->H3\n", "\n", "\n", "\n", "\n", "\n", "I13->H4\n", "\n", "\n", "\n", "\n", "\n", "I13->H5\n", "\n", "\n", "\n", "\n", "\n", "I13->H6\n", "\n", "\n", "\n", "\n", "\n", "I13->H7\n", "\n", "\n", "\n", "\n", "\n", "I13->H8\n", "\n", "\n", "\n", "\n", "\n", "I13->H9\n", "\n", "\n", "\n", "\n", "\n", "I13->H10\n", "\n", "\n", "\n", "\n", "\n", "I13->H11\n", "\n", "\n", "\n", "\n", "\n", "I13->H12\n", "\n", "\n", "\n", "\n", "\n", "I13->H13\n", "\n", "\n", "\n", "\n", "\n", "I13->H14\n", "\n", "\n", "\n", "\n", "\n", "I13->H15\n", "\n", "\n", "\n", "\n", "\n", "I13->H16\n", "\n", "\n", "\n", "\n", "\n", "I13->H17\n", "\n", "\n", "\n", "\n", "\n", "I13->H18\n", "\n", "\n", "\n", "\n", "\n", "I13->H19\n", "\n", "\n", "\n", "\n", "\n", "I13->H20\n", "\n", "\n", "\n", "\n", "\n", "I13->H21\n", "\n", "\n", "\n", "\n", "\n", "I13->H22\n", "\n", "\n", "\n", "\n", "\n", "I13->H23\n", "\n", "\n", "\n", "\n", "\n", "I13->H24\n", "\n", "\n", "\n", "\n", "\n", "I13->H25\n", "\n", "\n", "\n", "\n", "\n", "I13->H26\n", "\n", "\n", "\n", "\n", "\n", "I14\n", "\n", "Input 15\n", "\n", "\n", "\n", "I14->H0\n", "\n", "\n", "\n", "\n", "\n", "I14->H1\n", "\n", "\n", "\n", "\n", "\n", "I14->H2\n", "\n", "\n", "\n", "\n", "\n", "I14->H3\n", "\n", "\n", "\n", "\n", "\n", "I14->H4\n", "\n", "\n", "\n", "\n", "\n", "I14->H5\n", "\n", "\n", "\n", "\n", "\n", "I14->H6\n", "\n", "\n", "\n", "\n", "\n", "I14->H7\n", "\n", "\n", "\n", "\n", "\n", "I14->H8\n", "\n", "\n", "\n", "\n", "\n", "I14->H9\n", "\n", "\n", "\n", "\n", "\n", "I14->H10\n", "\n", "\n", "\n", "\n", "\n", "I14->H11\n", "\n", "\n", "\n", "\n", "\n", "I14->H12\n", "\n", "\n", "\n", "\n", "\n", "I14->H13\n", "\n", "\n", "\n", "\n", "\n", "I14->H14\n", "\n", "\n", "\n", "\n", "\n", "I14->H15\n", "\n", "\n", "\n", "\n", "\n", "I14->H16\n", "\n", "\n", "\n", "\n", "\n", "I14->H17\n", "\n", "\n", "\n", "\n", "\n", "I14->H18\n", "\n", "\n", "\n", "\n", "\n", "I14->H19\n", "\n", "\n", "\n", "\n", "\n", "I14->H20\n", "\n", "\n", "\n", "\n", "\n", "I14->H21\n", "\n", "\n", "\n", "\n", "\n", "I14->H22\n", "\n", "\n", "\n", "\n", "\n", "I14->H23\n", "\n", "\n", "\n", "\n", "\n", "I14->H24\n", "\n", "\n", "\n", "\n", "\n", "I14->H25\n", "\n", "\n", "\n", "\n", "\n", "I14->H26\n", "\n", "\n", "\n", "\n", "\n", "I15\n", "\n", "Input 16\n", "\n", "\n", "\n", "I15->H0\n", "\n", "\n", "\n", "\n", "\n", "I15->H1\n", "\n", "\n", "\n", "\n", "\n", "I15->H2\n", "\n", "\n", "\n", "\n", "\n", "I15->H3\n", "\n", "\n", "\n", "\n", "\n", "I15->H4\n", "\n", "\n", "\n", "\n", "\n", "I15->H5\n", "\n", "\n", "\n", "\n", "\n", "I15->H6\n", "\n", "\n", "\n", "\n", "\n", "I15->H7\n", "\n", "\n", "\n", "\n", "\n", "I15->H8\n", "\n", "\n", "\n", "\n", "\n", "I15->H9\n", "\n", "\n", "\n", "\n", "\n", "I15->H10\n", "\n", "\n", "\n", "\n", "\n", "I15->H11\n", "\n", "\n", "\n", "\n", "\n", "I15->H12\n", "\n", "\n", "\n", "\n", "\n", "I15->H13\n", "\n", "\n", "\n", "\n", "\n", "I15->H14\n", "\n", "\n", "\n", "\n", "\n", "I15->H15\n", "\n", "\n", "\n", "\n", "\n", "I15->H16\n", "\n", "\n", "\n", "\n", "\n", "I15->H17\n", "\n", "\n", "\n", "\n", "\n", "I15->H18\n", "\n", "\n", "\n", "\n", "\n", "I15->H19\n", "\n", "\n", "\n", "\n", "\n", "I15->H20\n", "\n", "\n", "\n", "\n", "\n", "I15->H21\n", "\n", "\n", "\n", "\n", "\n", "I15->H22\n", "\n", "\n", "\n", "\n", "\n", "I15->H23\n", "\n", "\n", "\n", "\n", "\n", "I15->H24\n", "\n", "\n", "\n", "\n", "\n", "I15->H25\n", "\n", "\n", "\n", "\n", "\n", "I15->H26\n", "\n", "\n", "\n", "\n", "\n", "I16\n", "\n", "Input 17\n", "\n", "\n", "\n", "I16->H0\n", "\n", "\n", "\n", "\n", "\n", "I16->H1\n", "\n", "\n", "\n", "\n", "\n", "I16->H2\n", "\n", "\n", "\n", "\n", "\n", "I16->H3\n", "\n", "\n", "\n", "\n", "\n", "I16->H4\n", "\n", "\n", "\n", "\n", "\n", "I16->H5\n", "\n", "\n", "\n", "\n", "\n", "I16->H6\n", "\n", "\n", "\n", "\n", "\n", "I16->H7\n", "\n", "\n", "\n", "\n", "\n", "I16->H8\n", "\n", "\n", "\n", "\n", "\n", "I16->H9\n", "\n", "\n", "\n", "\n", "\n", "I16->H10\n", "\n", "\n", "\n", "\n", "\n", "I16->H11\n", "\n", "\n", "\n", "\n", "\n", "I16->H12\n", "\n", "\n", "\n", "\n", "\n", "I16->H13\n", "\n", "\n", "\n", "\n", "\n", "I16->H14\n", "\n", "\n", "\n", "\n", "\n", "I16->H15\n", "\n", "\n", "\n", "\n", "\n", "I16->H16\n", "\n", "\n", "\n", "\n", "\n", "I16->H17\n", "\n", "\n", "\n", "\n", "\n", "I16->H18\n", "\n", "\n", "\n", "\n", "\n", "I16->H19\n", "\n", "\n", "\n", "\n", "\n", "I16->H20\n", "\n", "\n", "\n", "\n", "\n", "I16->H21\n", "\n", "\n", "\n", "\n", "\n", "I16->H22\n", "\n", "\n", "\n", "\n", "\n", "I16->H23\n", "\n", "\n", "\n", "\n", "\n", "I16->H24\n", "\n", "\n", "\n", "\n", "\n", "I16->H25\n", "\n", "\n", "\n", "\n", "\n", "I16->H26\n", "\n", "\n", "\n", "\n", "\n", "I17\n", "\n", "Input 18\n", "\n", "\n", "\n", "I17->H0\n", "\n", "\n", "\n", "\n", "\n", "I17->H1\n", "\n", "\n", "\n", "\n", "\n", "I17->H2\n", "\n", "\n", "\n", "\n", "\n", "I17->H3\n", "\n", "\n", "\n", "\n", "\n", "I17->H4\n", "\n", "\n", "\n", "\n", "\n", "I17->H5\n", "\n", "\n", "\n", "\n", "\n", "I17->H6\n", "\n", "\n", "\n", "\n", "\n", "I17->H7\n", "\n", "\n", "\n", "\n", "\n", "I17->H8\n", "\n", "\n", "\n", "\n", "\n", "I17->H9\n", "\n", "\n", "\n", "\n", "\n", "I17->H10\n", "\n", "\n", "\n", "\n", "\n", "I17->H11\n", "\n", "\n", "\n", "\n", "\n", "I17->H12\n", "\n", "\n", "\n", "\n", "\n", "I17->H13\n", "\n", "\n", "\n", "\n", "\n", "I17->H14\n", "\n", "\n", "\n", "\n", "\n", "I17->H15\n", "\n", "\n", "\n", "\n", "\n", "I17->H16\n", "\n", "\n", "\n", "\n", "\n", "I17->H17\n", "\n", "\n", "\n", "\n", "\n", "I17->H18\n", "\n", "\n", "\n", "\n", "\n", "I17->H19\n", "\n", "\n", "\n", "\n", "\n", "I17->H20\n", "\n", "\n", "\n", "\n", "\n", "I17->H21\n", "\n", "\n", "\n", "\n", "\n", "I17->H22\n", "\n", "\n", "\n", "\n", "\n", "I17->H23\n", "\n", "\n", "\n", "\n", "\n", "I17->H24\n", "\n", "\n", "\n", "\n", "\n", "I17->H25\n", "\n", "\n", "\n", "\n", "\n", "I17->H26\n", "\n", "\n", "\n", "\n", "\n", "I18\n", "\n", "Input 19\n", "\n", "\n", "\n", "I18->H0\n", "\n", "\n", "\n", "\n", "\n", "I18->H1\n", "\n", "\n", "\n", "\n", "\n", "I18->H2\n", "\n", "\n", "\n", "\n", "\n", "I18->H3\n", "\n", "\n", "\n", "\n", "\n", "I18->H4\n", "\n", "\n", "\n", "\n", "\n", "I18->H5\n", "\n", "\n", "\n", "\n", "\n", "I18->H6\n", "\n", "\n", "\n", "\n", "\n", "I18->H7\n", "\n", "\n", "\n", "\n", "\n", "I18->H8\n", "\n", "\n", "\n", "\n", "\n", "I18->H9\n", "\n", "\n", "\n", "\n", "\n", "I18->H10\n", "\n", "\n", "\n", "\n", "\n", "I18->H11\n", "\n", "\n", "\n", "\n", "\n", "I18->H12\n", "\n", "\n", "\n", "\n", "\n", "I18->H13\n", "\n", "\n", "\n", "\n", "\n", "I18->H14\n", "\n", "\n", "\n", "\n", "\n", "I18->H15\n", "\n", "\n", "\n", "\n", "\n", "I18->H16\n", "\n", "\n", "\n", "\n", "\n", "I18->H17\n", "\n", "\n", "\n", "\n", "\n", "I18->H18\n", "\n", "\n", "\n", "\n", "\n", "I18->H19\n", "\n", "\n", "\n", "\n", "\n", "I18->H20\n", "\n", "\n", "\n", "\n", "\n", "I18->H21\n", "\n", "\n", "\n", "\n", "\n", "I18->H22\n", "\n", "\n", "\n", "\n", "\n", "I18->H23\n", "\n", "\n", "\n", "\n", "\n", "I18->H24\n", "\n", "\n", "\n", "\n", "\n", "I18->H25\n", "\n", "\n", "\n", "\n", "\n", "I18->H26\n", "\n", "\n", "\n", "\n", "\n", "I19\n", "\n", "Input 20\n", "\n", "\n", "\n", "I19->H0\n", "\n", "\n", "\n", "\n", "\n", "I19->H1\n", "\n", "\n", "\n", "\n", "\n", "I19->H2\n", "\n", "\n", "\n", "\n", "\n", "I19->H3\n", "\n", "\n", "\n", "\n", "\n", "I19->H4\n", "\n", "\n", "\n", "\n", "\n", "I19->H5\n", "\n", "\n", "\n", "\n", "\n", "I19->H6\n", "\n", "\n", "\n", "\n", "\n", "I19->H7\n", "\n", "\n", "\n", "\n", "\n", "I19->H8\n", "\n", "\n", "\n", "\n", "\n", "I19->H9\n", "\n", "\n", "\n", "\n", "\n", "I19->H10\n", "\n", "\n", "\n", "\n", "\n", "I19->H11\n", "\n", "\n", "\n", "\n", "\n", "I19->H12\n", "\n", "\n", "\n", "\n", "\n", "I19->H13\n", "\n", "\n", "\n", "\n", "\n", "I19->H14\n", "\n", "\n", "\n", "\n", "\n", "I19->H15\n", "\n", "\n", "\n", "\n", "\n", "I19->H16\n", "\n", "\n", "\n", "\n", "\n", "I19->H17\n", "\n", "\n", "\n", "\n", "\n", "I19->H18\n", "\n", "\n", "\n", "\n", "\n", "I19->H19\n", "\n", "\n", "\n", "\n", "\n", "I19->H20\n", "\n", "\n", "\n", "\n", "\n", "I19->H21\n", "\n", "\n", "\n", "\n", "\n", "I19->H22\n", "\n", "\n", "\n", "\n", "\n", "I19->H23\n", "\n", "\n", "\n", "\n", "\n", "I19->H24\n", "\n", "\n", "\n", "\n", "\n", "I19->H25\n", "\n", "\n", "\n", "\n", "\n", "I19->H26\n", "\n", "\n", "\n", "\n", "\n", "I20\n", "\n", "Input 21\n", "\n", "\n", "\n", "I20->H0\n", "\n", "\n", "\n", "\n", "\n", "I20->H1\n", "\n", "\n", "\n", "\n", "\n", "I20->H2\n", "\n", "\n", "\n", "\n", "\n", "I20->H3\n", "\n", "\n", "\n", "\n", "\n", "I20->H4\n", "\n", "\n", "\n", "\n", "\n", "I20->H5\n", "\n", "\n", "\n", "\n", "\n", "I20->H6\n", "\n", "\n", "\n", "\n", "\n", "I20->H7\n", "\n", "\n", "\n", "\n", "\n", "I20->H8\n", "\n", "\n", "\n", "\n", "\n", "I20->H9\n", "\n", "\n", "\n", "\n", "\n", "I20->H10\n", "\n", "\n", "\n", "\n", "\n", "I20->H11\n", "\n", "\n", "\n", "\n", "\n", "I20->H12\n", "\n", "\n", "\n", "\n", "\n", "I20->H13\n", "\n", "\n", "\n", "\n", "\n", "I20->H14\n", "\n", "\n", "\n", "\n", "\n", "I20->H15\n", "\n", "\n", "\n", "\n", "\n", "I20->H16\n", "\n", "\n", "\n", "\n", "\n", "I20->H17\n", "\n", "\n", "\n", "\n", "\n", "I20->H18\n", "\n", "\n", "\n", "\n", "\n", "I20->H19\n", "\n", "\n", "\n", "\n", "\n", "I20->H20\n", "\n", "\n", "\n", "\n", "\n", "I20->H21\n", "\n", "\n", "\n", "\n", "\n", "I20->H22\n", "\n", "\n", "\n", "\n", "\n", "I20->H23\n", "\n", "\n", "\n", "\n", "\n", "I20->H24\n", "\n", "\n", "\n", "\n", "\n", "I20->H25\n", "\n", "\n", "\n", "\n", "\n", "I20->H26\n", "\n", "\n", "\n", "\n", "\n", "I21\n", "\n", "Input 22\n", "\n", "\n", "\n", "I21->H0\n", "\n", "\n", "\n", "\n", "\n", "I21->H1\n", "\n", "\n", "\n", "\n", "\n", "I21->H2\n", "\n", "\n", "\n", "\n", "\n", "I21->H3\n", "\n", "\n", "\n", "\n", "\n", "I21->H4\n", "\n", "\n", "\n", "\n", "\n", "I21->H5\n", "\n", "\n", "\n", "\n", "\n", "I21->H6\n", "\n", "\n", "\n", "\n", "\n", "I21->H7\n", "\n", "\n", "\n", "\n", "\n", "I21->H8\n", "\n", "\n", "\n", "\n", "\n", "I21->H9\n", "\n", "\n", "\n", "\n", "\n", "I21->H10\n", "\n", "\n", "\n", "\n", "\n", "I21->H11\n", "\n", "\n", "\n", "\n", "\n", "I21->H12\n", "\n", "\n", "\n", "\n", "\n", "I21->H13\n", "\n", "\n", "\n", "\n", "\n", "I21->H14\n", "\n", "\n", "\n", "\n", "\n", "I21->H15\n", "\n", "\n", "\n", "\n", "\n", "I21->H16\n", "\n", "\n", "\n", "\n", "\n", "I21->H17\n", "\n", "\n", "\n", "\n", "\n", "I21->H18\n", "\n", "\n", "\n", "\n", "\n", "I21->H19\n", "\n", "\n", "\n", "\n", "\n", "I21->H20\n", "\n", "\n", "\n", "\n", "\n", "I21->H21\n", "\n", "\n", "\n", "\n", "\n", "I21->H22\n", "\n", "\n", "\n", "\n", "\n", "I21->H23\n", "\n", "\n", "\n", "\n", "\n", "I21->H24\n", "\n", "\n", "\n", "\n", "\n", "I21->H25\n", "\n", "\n", "\n", "\n", "\n", "I21->H26\n", "\n", "\n", "\n", "\n", "\n", "I22\n", "\n", "Input 23\n", "\n", "\n", "\n", "I22->H0\n", "\n", "\n", "\n", "\n", "\n", "I22->H1\n", "\n", "\n", "\n", "\n", "\n", "I22->H2\n", "\n", "\n", "\n", "\n", "\n", "I22->H3\n", "\n", "\n", "\n", "\n", "\n", "I22->H4\n", "\n", "\n", "\n", "\n", "\n", "I22->H5\n", "\n", "\n", "\n", "\n", "\n", "I22->H6\n", "\n", "\n", "\n", "\n", "\n", "I22->H7\n", "\n", "\n", "\n", "\n", "\n", "I22->H8\n", "\n", "\n", "\n", "\n", "\n", "I22->H9\n", "\n", "\n", "\n", "\n", "\n", "I22->H10\n", "\n", "\n", "\n", "\n", "\n", "I22->H11\n", "\n", "\n", "\n", "\n", "\n", "I22->H12\n", "\n", "\n", "\n", "\n", "\n", "I22->H13\n", "\n", "\n", "\n", "\n", "\n", "I22->H14\n", "\n", "\n", "\n", "\n", "\n", "I22->H15\n", "\n", "\n", "\n", "\n", "\n", "I22->H16\n", "\n", "\n", "\n", "\n", "\n", "I22->H17\n", "\n", "\n", "\n", "\n", "\n", "I22->H18\n", "\n", "\n", "\n", "\n", "\n", "I22->H19\n", "\n", "\n", "\n", "\n", "\n", "I22->H20\n", "\n", "\n", "\n", "\n", "\n", "I22->H21\n", "\n", "\n", "\n", "\n", "\n", "I22->H22\n", "\n", "\n", "\n", "\n", "\n", "I22->H23\n", "\n", "\n", "\n", "\n", "\n", "I22->H24\n", "\n", "\n", "\n", "\n", "\n", "I22->H25\n", "\n", "\n", "\n", "\n", "\n", "I22->H26\n", "\n", "\n", "\n", "\n", "\n", "I23\n", "\n", "Input 24\n", "\n", "\n", "\n", "I23->H0\n", "\n", "\n", "\n", "\n", "\n", "I23->H1\n", "\n", "\n", "\n", "\n", "\n", "I23->H2\n", "\n", "\n", "\n", "\n", "\n", "I23->H3\n", "\n", "\n", "\n", "\n", "\n", "I23->H4\n", "\n", "\n", "\n", "\n", "\n", "I23->H5\n", "\n", "\n", "\n", "\n", "\n", "I23->H6\n", "\n", "\n", "\n", "\n", "\n", "I23->H7\n", "\n", "\n", "\n", "\n", "\n", "I23->H8\n", "\n", "\n", "\n", "\n", "\n", "I23->H9\n", "\n", "\n", "\n", "\n", "\n", "I23->H10\n", "\n", "\n", "\n", "\n", "\n", "I23->H11\n", "\n", "\n", "\n", "\n", "\n", "I23->H12\n", "\n", "\n", "\n", "\n", "\n", "I23->H13\n", "\n", "\n", "\n", "\n", "\n", "I23->H14\n", "\n", "\n", "\n", "\n", "\n", "I23->H15\n", "\n", "\n", "\n", "\n", "\n", "I23->H16\n", "\n", "\n", "\n", "\n", "\n", "I23->H17\n", "\n", "\n", "\n", "\n", "\n", "I23->H18\n", "\n", "\n", "\n", "\n", "\n", "I23->H19\n", "\n", "\n", "\n", "\n", "\n", "I23->H20\n", "\n", "\n", "\n", "\n", "\n", "I23->H21\n", "\n", "\n", "\n", "\n", "\n", "I23->H22\n", "\n", "\n", "\n", "\n", "\n", "I23->H23\n", "\n", "\n", "\n", "\n", "\n", "I23->H24\n", "\n", "\n", "\n", "\n", "\n", "I23->H25\n", "\n", "\n", "\n", "\n", "\n", "I23->H26\n", "\n", "\n", "\n", "\n", "\n", "I24\n", "\n", "Input 25\n", "\n", "\n", "\n", "I24->H0\n", "\n", "\n", "\n", "\n", "\n", "I24->H1\n", "\n", "\n", "\n", "\n", "\n", "I24->H2\n", "\n", "\n", "\n", "\n", "\n", "I24->H3\n", "\n", "\n", "\n", "\n", "\n", "I24->H4\n", "\n", "\n", "\n", "\n", "\n", "I24->H5\n", "\n", "\n", "\n", "\n", "\n", "I24->H6\n", "\n", "\n", "\n", "\n", "\n", "I24->H7\n", "\n", "\n", "\n", "\n", "\n", "I24->H8\n", "\n", "\n", "\n", "\n", "\n", "I24->H9\n", "\n", "\n", "\n", "\n", "\n", "I24->H10\n", "\n", "\n", "\n", "\n", "\n", "I24->H11\n", "\n", "\n", "\n", "\n", "\n", "I24->H12\n", "\n", "\n", "\n", "\n", "\n", "I24->H13\n", "\n", "\n", "\n", "\n", "\n", "I24->H14\n", "\n", "\n", "\n", "\n", "\n", "I24->H15\n", "\n", "\n", "\n", "\n", "\n", "I24->H16\n", "\n", "\n", "\n", "\n", "\n", "I24->H17\n", "\n", "\n", "\n", "\n", "\n", "I24->H18\n", "\n", "\n", "\n", "\n", "\n", "I24->H19\n", "\n", "\n", "\n", "\n", "\n", "I24->H20\n", "\n", "\n", "\n", "\n", "\n", "I24->H21\n", "\n", "\n", "\n", "\n", "\n", "I24->H22\n", "\n", "\n", "\n", "\n", "\n", "I24->H23\n", "\n", "\n", "\n", "\n", "\n", "I24->H24\n", "\n", "\n", "\n", "\n", "\n", "I24->H25\n", "\n", "\n", "\n", "\n", "\n", "I24->H26\n", "\n", "\n", "\n", "\n", "\n", "I25\n", "\n", "Input 26\n", "\n", "\n", "\n", "I25->H0\n", "\n", "\n", "\n", "\n", "\n", "I25->H1\n", "\n", "\n", "\n", "\n", "\n", "I25->H2\n", "\n", "\n", "\n", "\n", "\n", "I25->H3\n", "\n", "\n", "\n", "\n", "\n", "I25->H4\n", "\n", "\n", "\n", "\n", "\n", "I25->H5\n", "\n", "\n", "\n", "\n", "\n", "I25->H6\n", "\n", "\n", "\n", "\n", "\n", "I25->H7\n", "\n", "\n", "\n", "\n", "\n", "I25->H8\n", "\n", "\n", "\n", "\n", "\n", "I25->H9\n", "\n", "\n", "\n", "\n", "\n", "I25->H10\n", "\n", "\n", "\n", "\n", "\n", "I25->H11\n", "\n", "\n", "\n", "\n", "\n", "I25->H12\n", "\n", "\n", "\n", "\n", "\n", "I25->H13\n", "\n", "\n", "\n", "\n", "\n", "I25->H14\n", "\n", "\n", "\n", "\n", "\n", "I25->H15\n", "\n", "\n", "\n", "\n", "\n", "I25->H16\n", "\n", "\n", "\n", "\n", "\n", "I25->H17\n", "\n", "\n", "\n", "\n", "\n", "I25->H18\n", "\n", "\n", "\n", "\n", "\n", "I25->H19\n", "\n", "\n", "\n", "\n", "\n", "I25->H20\n", "\n", "\n", "\n", "\n", "\n", "I25->H21\n", "\n", "\n", "\n", "\n", "\n", "I25->H22\n", "\n", "\n", "\n", "\n", "\n", "I25->H23\n", "\n", "\n", "\n", "\n", "\n", "I25->H24\n", "\n", "\n", "\n", "\n", "\n", "I25->H25\n", "\n", "\n", "\n", "\n", "\n", "I25->H26\n", "\n", "\n", "\n", "\n", "\n", "I26\n", "\n", "Input 27\n", "\n", "\n", "\n", "I26->H0\n", "\n", "\n", "\n", "\n", "\n", "I26->H1\n", "\n", "\n", "\n", "\n", "\n", "I26->H2\n", "\n", "\n", "\n", "\n", "\n", "I26->H3\n", "\n", "\n", "\n", "\n", "\n", "I26->H4\n", "\n", "\n", "\n", "\n", "\n", "I26->H5\n", "\n", "\n", "\n", "\n", "\n", "I26->H6\n", "\n", "\n", "\n", "\n", "\n", "I26->H7\n", "\n", "\n", "\n", "\n", "\n", "I26->H8\n", "\n", "\n", "\n", "\n", "\n", "I26->H9\n", "\n", "\n", "\n", "\n", "\n", "I26->H10\n", "\n", "\n", "\n", "\n", "\n", "I26->H11\n", "\n", "\n", "\n", "\n", "\n", "I26->H12\n", "\n", "\n", "\n", "\n", "\n", "I26->H13\n", "\n", "\n", "\n", "\n", "\n", "I26->H14\n", "\n", "\n", "\n", "\n", "\n", "I26->H15\n", "\n", "\n", "\n", "\n", "\n", "I26->H16\n", "\n", "\n", "\n", "\n", "\n", "I26->H17\n", "\n", "\n", "\n", "\n", "\n", "I26->H18\n", "\n", "\n", "\n", "\n", "\n", "I26->H19\n", "\n", "\n", "\n", "\n", "\n", "I26->H20\n", "\n", "\n", "\n", "\n", "\n", "I26->H21\n", "\n", "\n", "\n", "\n", "\n", "I26->H22\n", "\n", "\n", "\n", "\n", "\n", "I26->H23\n", "\n", "\n", "\n", "\n", "\n", "I26->H24\n", "\n", "\n", "\n", "\n", "\n", "I26->H25\n", "\n", "\n", "\n", "\n", "\n", "I26->H26\n", "\n", "\n", "\n", "\n", "\n", "O0\n", "\n", "Softmax 1\n", "\n", "\n", "\n", "H0->O0\n", "\n", "\n", "\n", "\n", "\n", "O1\n", "\n", "Softmax 2\n", "\n", "\n", "\n", "H0->O1\n", "\n", "\n", "\n", "\n", "\n", "O2\n", "\n", "Softmax 3\n", "\n", "\n", "\n", "H0->O2\n", "\n", "\n", "\n", "\n", "\n", "O3\n", "\n", "Softmax 4\n", "\n", "\n", "\n", "H0->O3\n", "\n", "\n", "\n", "\n", "\n", "O4\n", "\n", "Softmax 5\n", "\n", "\n", "\n", "H0->O4\n", "\n", "\n", "\n", "\n", "\n", "O5\n", "\n", "Softmax 6\n", "\n", "\n", "\n", "H0->O5\n", "\n", "\n", "\n", "\n", "\n", "O6\n", "\n", "Softmax 7\n", "\n", "\n", "\n", "H0->O6\n", "\n", "\n", "\n", "\n", "\n", "O7\n", "\n", "Softmax 8\n", "\n", "\n", "\n", "H0->O7\n", "\n", "\n", "\n", "\n", "\n", "O8\n", "\n", "Softmax 9\n", "\n", "\n", "\n", "H0->O8\n", "\n", "\n", "\n", "\n", "\n", "O9\n", "\n", "Softmax 10\n", "\n", "\n", "\n", "H0->O9\n", "\n", "\n", "\n", "\n", "\n", "O10\n", "\n", "Softmax 11\n", "\n", "\n", "\n", "H0->O10\n", "\n", "\n", "\n", "\n", "\n", "O11\n", "\n", "Softmax 12\n", "\n", "\n", "\n", "H0->O11\n", "\n", "\n", "\n", "\n", "\n", "O12\n", "\n", "Softmax 13\n", "\n", "\n", "\n", "H0->O12\n", "\n", "\n", "\n", "\n", "\n", "O13\n", "\n", "Softmax 14\n", "\n", "\n", "\n", "H0->O13\n", "\n", "\n", "\n", "\n", "\n", "O14\n", "\n", "Softmax 15\n", "\n", "\n", "\n", "H0->O14\n", "\n", "\n", "\n", "\n", "\n", "O15\n", "\n", "Softmax 16\n", "\n", "\n", "\n", "H0->O15\n", "\n", "\n", "\n", "\n", "\n", "O16\n", "\n", "Softmax 17\n", "\n", "\n", "\n", "H0->O16\n", "\n", "\n", "\n", "\n", "\n", "O17\n", "\n", "Softmax 18\n", "\n", "\n", "\n", "H0->O17\n", "\n", "\n", "\n", "\n", "\n", "O18\n", "\n", "Softmax 19\n", "\n", "\n", "\n", "H0->O18\n", "\n", "\n", "\n", "\n", "\n", "O19\n", "\n", "Softmax 20\n", "\n", "\n", "\n", "H0->O19\n", "\n", "\n", "\n", "\n", "\n", "O20\n", "\n", "Softmax 21\n", "\n", "\n", "\n", "H0->O20\n", "\n", "\n", "\n", "\n", "\n", "O21\n", "\n", "Softmax 22\n", "\n", "\n", "\n", "H0->O21\n", "\n", "\n", "\n", "\n", "\n", "O22\n", "\n", "Softmax 23\n", "\n", "\n", "\n", "H0->O22\n", "\n", "\n", "\n", "\n", "\n", "O23\n", "\n", "Softmax 24\n", "\n", "\n", "\n", "H0->O23\n", "\n", "\n", "\n", "\n", "\n", "O24\n", "\n", "Softmax 25\n", "\n", "\n", "\n", "H0->O24\n", "\n", "\n", "\n", "\n", "\n", "O25\n", "\n", "Softmax 26\n", "\n", "\n", "\n", "H0->O25\n", "\n", "\n", "\n", "\n", "\n", "O26\n", "\n", "Softmax 27\n", "\n", "\n", "\n", "H0->O26\n", "\n", "\n", "\n", "\n", "\n", "H1->O0\n", "\n", "\n", "\n", "\n", "\n", "H1->O1\n", "\n", "\n", "\n", "\n", "\n", "H1->O2\n", "\n", "\n", "\n", "\n", "\n", "H1->O3\n", "\n", "\n", "\n", "\n", "\n", "H1->O4\n", "\n", "\n", "\n", "\n", "\n", "H1->O5\n", "\n", "\n", "\n", "\n", "\n", "H1->O6\n", "\n", "\n", "\n", "\n", "\n", "H1->O7\n", "\n", "\n", "\n", "\n", "\n", "H1->O8\n", "\n", "\n", "\n", "\n", "\n", "H1->O9\n", "\n", "\n", "\n", "\n", "\n", "H1->O10\n", "\n", "\n", "\n", "\n", "\n", "H1->O11\n", "\n", "\n", "\n", "\n", "\n", "H1->O12\n", "\n", "\n", "\n", "\n", "\n", "H1->O13\n", "\n", "\n", "\n", "\n", "\n", "H1->O14\n", "\n", "\n", "\n", "\n", "\n", "H1->O15\n", "\n", "\n", "\n", "\n", "\n", "H1->O16\n", "\n", "\n", "\n", "\n", "\n", "H1->O17\n", "\n", "\n", "\n", "\n", "\n", "H1->O18\n", "\n", "\n", "\n", "\n", "\n", "H1->O19\n", "\n", "\n", "\n", "\n", "\n", "H1->O20\n", "\n", "\n", "\n", "\n", "\n", "H1->O21\n", "\n", "\n", "\n", "\n", "\n", "H1->O22\n", "\n", "\n", "\n", "\n", "\n", "H1->O23\n", "\n", "\n", "\n", "\n", "\n", "H1->O24\n", "\n", "\n", "\n", "\n", "\n", "H1->O25\n", "\n", "\n", "\n", "\n", "\n", "H1->O26\n", "\n", "\n", "\n", "\n", "\n", "H2->O0\n", "\n", "\n", "\n", "\n", "\n", "H2->O1\n", "\n", "\n", "\n", "\n", "\n", "H2->O2\n", "\n", "\n", "\n", "\n", "\n", "H2->O3\n", "\n", "\n", "\n", "\n", "\n", "H2->O4\n", "\n", "\n", "\n", "\n", "\n", "H2->O5\n", "\n", "\n", "\n", "\n", "\n", "H2->O6\n", "\n", "\n", "\n", "\n", "\n", "H2->O7\n", "\n", "\n", "\n", "\n", "\n", "H2->O8\n", "\n", "\n", "\n", "\n", "\n", "H2->O9\n", "\n", "\n", "\n", "\n", "\n", "H2->O10\n", "\n", "\n", "\n", "\n", "\n", "H2->O11\n", "\n", "\n", "\n", "\n", "\n", "H2->O12\n", "\n", "\n", "\n", "\n", "\n", "H2->O13\n", "\n", "\n", "\n", "\n", "\n", "H2->O14\n", "\n", "\n", "\n", "\n", "\n", "H2->O15\n", "\n", "\n", "\n", "\n", "\n", "H2->O16\n", "\n", "\n", "\n", "\n", "\n", "H2->O17\n", "\n", "\n", "\n", "\n", "\n", "H2->O18\n", "\n", "\n", "\n", "\n", "\n", "H2->O19\n", "\n", "\n", "\n", "\n", "\n", "H2->O20\n", "\n", "\n", "\n", "\n", "\n", "H2->O21\n", "\n", "\n", "\n", "\n", "\n", "H2->O22\n", "\n", "\n", "\n", "\n", "\n", "H2->O23\n", "\n", "\n", "\n", "\n", "\n", "H2->O24\n", "\n", "\n", "\n", "\n", "\n", "H2->O25\n", "\n", "\n", "\n", "\n", "\n", "H2->O26\n", "\n", "\n", "\n", "\n", "\n", "H3->O0\n", "\n", "\n", "\n", "\n", "\n", "H3->O1\n", "\n", "\n", "\n", "\n", "\n", "H3->O2\n", "\n", "\n", "\n", "\n", "\n", "H3->O3\n", "\n", "\n", "\n", "\n", "\n", "H3->O4\n", "\n", "\n", "\n", "\n", "\n", "H3->O5\n", "\n", "\n", "\n", "\n", "\n", "H3->O6\n", "\n", "\n", "\n", "\n", "\n", "H3->O7\n", "\n", "\n", "\n", "\n", "\n", "H3->O8\n", "\n", "\n", "\n", "\n", "\n", "H3->O9\n", "\n", "\n", "\n", "\n", "\n", "H3->O10\n", "\n", "\n", "\n", "\n", "\n", "H3->O11\n", "\n", "\n", "\n", "\n", "\n", "H3->O12\n", "\n", "\n", "\n", "\n", "\n", "H3->O13\n", "\n", "\n", "\n", "\n", "\n", "H3->O14\n", "\n", "\n", "\n", "\n", "\n", "H3->O15\n", "\n", "\n", "\n", "\n", "\n", "H3->O16\n", "\n", "\n", "\n", "\n", "\n", "H3->O17\n", "\n", "\n", "\n", "\n", "\n", "H3->O18\n", "\n", "\n", "\n", "\n", "\n", "H3->O19\n", "\n", "\n", "\n", "\n", "\n", "H3->O20\n", "\n", "\n", "\n", "\n", "\n", "H3->O21\n", "\n", "\n", "\n", "\n", "\n", "H3->O22\n", "\n", "\n", "\n", "\n", "\n", "H3->O23\n", "\n", "\n", "\n", "\n", "\n", "H3->O24\n", "\n", "\n", "\n", "\n", "\n", "H3->O25\n", "\n", "\n", "\n", "\n", "\n", "H3->O26\n", "\n", "\n", "\n", "\n", "\n", "H4->O0\n", "\n", "\n", "\n", "\n", "\n", "H4->O1\n", "\n", "\n", "\n", "\n", "\n", "H4->O2\n", "\n", "\n", "\n", "\n", "\n", "H4->O3\n", "\n", "\n", "\n", "\n", "\n", "H4->O4\n", "\n", "\n", "\n", "\n", "\n", "H4->O5\n", "\n", "\n", "\n", "\n", "\n", "H4->O6\n", "\n", "\n", "\n", "\n", "\n", "H4->O7\n", "\n", "\n", "\n", "\n", "\n", "H4->O8\n", "\n", "\n", "\n", "\n", "\n", "H4->O9\n", "\n", "\n", "\n", "\n", "\n", "H4->O10\n", "\n", "\n", "\n", "\n", "\n", "H4->O11\n", "\n", "\n", "\n", "\n", "\n", "H4->O12\n", "\n", "\n", "\n", "\n", "\n", "H4->O13\n", "\n", "\n", "\n", "\n", "\n", "H4->O14\n", "\n", "\n", "\n", "\n", "\n", "H4->O15\n", "\n", "\n", "\n", "\n", "\n", "H4->O16\n", "\n", "\n", "\n", "\n", "\n", "H4->O17\n", "\n", "\n", "\n", "\n", "\n", "H4->O18\n", "\n", "\n", "\n", "\n", "\n", "H4->O19\n", "\n", "\n", "\n", "\n", "\n", "H4->O20\n", "\n", "\n", "\n", "\n", "\n", "H4->O21\n", "\n", "\n", "\n", "\n", "\n", "H4->O22\n", "\n", "\n", "\n", "\n", "\n", "H4->O23\n", "\n", "\n", "\n", "\n", "\n", "H4->O24\n", "\n", "\n", "\n", "\n", "\n", "H4->O25\n", "\n", "\n", "\n", "\n", "\n", "H4->O26\n", "\n", "\n", "\n", "\n", "\n", "H5->O0\n", "\n", "\n", "\n", "\n", "\n", "H5->O1\n", "\n", "\n", "\n", "\n", "\n", "H5->O2\n", "\n", "\n", "\n", "\n", "\n", "H5->O3\n", "\n", "\n", "\n", "\n", "\n", "H5->O4\n", "\n", "\n", "\n", "\n", "\n", "H5->O5\n", "\n", "\n", "\n", "\n", "\n", "H5->O6\n", "\n", "\n", "\n", "\n", "\n", "H5->O7\n", "\n", "\n", "\n", "\n", "\n", "H5->O8\n", "\n", "\n", "\n", "\n", "\n", "H5->O9\n", "\n", "\n", "\n", "\n", "\n", "H5->O10\n", "\n", "\n", "\n", "\n", "\n", "H5->O11\n", "\n", "\n", "\n", "\n", "\n", "H5->O12\n", "\n", "\n", "\n", "\n", "\n", "H5->O13\n", "\n", "\n", "\n", "\n", "\n", "H5->O14\n", "\n", "\n", "\n", "\n", "\n", "H5->O15\n", "\n", "\n", "\n", "\n", "\n", "H5->O16\n", "\n", "\n", "\n", "\n", "\n", "H5->O17\n", "\n", "\n", "\n", "\n", "\n", "H5->O18\n", "\n", "\n", "\n", "\n", "\n", "H5->O19\n", "\n", "\n", "\n", "\n", "\n", "H5->O20\n", "\n", "\n", "\n", "\n", "\n", "H5->O21\n", "\n", "\n", "\n", "\n", "\n", "H5->O22\n", "\n", "\n", "\n", "\n", "\n", "H5->O23\n", "\n", "\n", "\n", "\n", "\n", "H5->O24\n", "\n", "\n", "\n", "\n", "\n", "H5->O25\n", "\n", "\n", "\n", "\n", "\n", "H5->O26\n", "\n", "\n", "\n", "\n", "\n", "H6->O0\n", "\n", "\n", "\n", "\n", "\n", "H6->O1\n", "\n", "\n", "\n", "\n", "\n", "H6->O2\n", "\n", "\n", "\n", "\n", "\n", "H6->O3\n", "\n", "\n", "\n", "\n", "\n", "H6->O4\n", "\n", "\n", "\n", "\n", "\n", "H6->O5\n", "\n", "\n", "\n", "\n", "\n", "H6->O6\n", "\n", "\n", "\n", "\n", "\n", "H6->O7\n", "\n", "\n", "\n", "\n", "\n", "H6->O8\n", "\n", "\n", "\n", "\n", "\n", "H6->O9\n", "\n", "\n", "\n", "\n", "\n", "H6->O10\n", "\n", "\n", "\n", "\n", "\n", "H6->O11\n", "\n", "\n", "\n", "\n", "\n", "H6->O12\n", "\n", "\n", "\n", "\n", "\n", "H6->O13\n", "\n", "\n", "\n", "\n", "\n", "H6->O14\n", "\n", "\n", "\n", "\n", "\n", "H6->O15\n", "\n", "\n", "\n", "\n", "\n", "H6->O16\n", "\n", "\n", "\n", "\n", "\n", "H6->O17\n", "\n", "\n", "\n", "\n", "\n", "H6->O18\n", "\n", "\n", "\n", "\n", "\n", "H6->O19\n", "\n", "\n", "\n", "\n", "\n", "H6->O20\n", "\n", "\n", "\n", "\n", "\n", "H6->O21\n", "\n", "\n", "\n", "\n", "\n", "H6->O22\n", "\n", "\n", "\n", "\n", "\n", "H6->O23\n", "\n", "\n", "\n", "\n", "\n", "H6->O24\n", "\n", "\n", "\n", "\n", "\n", "H6->O25\n", "\n", "\n", "\n", "\n", "\n", "H6->O26\n", "\n", "\n", "\n", "\n", "\n", "H7->O0\n", "\n", "\n", "\n", "\n", "\n", "H7->O1\n", "\n", "\n", "\n", "\n", "\n", "H7->O2\n", "\n", "\n", "\n", "\n", "\n", "H7->O3\n", "\n", "\n", "\n", "\n", "\n", "H7->O4\n", "\n", "\n", "\n", "\n", "\n", "H7->O5\n", "\n", "\n", "\n", "\n", "\n", "H7->O6\n", "\n", "\n", "\n", "\n", "\n", "H7->O7\n", "\n", "\n", "\n", "\n", "\n", "H7->O8\n", "\n", "\n", "\n", "\n", "\n", "H7->O9\n", "\n", "\n", "\n", "\n", "\n", "H7->O10\n", "\n", "\n", "\n", "\n", "\n", "H7->O11\n", "\n", "\n", "\n", "\n", "\n", "H7->O12\n", "\n", "\n", "\n", "\n", "\n", "H7->O13\n", "\n", "\n", "\n", "\n", "\n", "H7->O14\n", "\n", "\n", "\n", "\n", "\n", "H7->O15\n", "\n", "\n", "\n", "\n", "\n", "H7->O16\n", "\n", "\n", "\n", "\n", "\n", "H7->O17\n", "\n", "\n", "\n", "\n", "\n", "H7->O18\n", "\n", "\n", "\n", "\n", "\n", "H7->O19\n", "\n", "\n", "\n", "\n", "\n", "H7->O20\n", "\n", "\n", "\n", "\n", "\n", "H7->O21\n", "\n", "\n", "\n", "\n", "\n", "H7->O22\n", "\n", "\n", "\n", "\n", "\n", "H7->O23\n", "\n", "\n", "\n", "\n", "\n", "H7->O24\n", "\n", "\n", "\n", "\n", "\n", "H7->O25\n", "\n", "\n", "\n", "\n", "\n", "H7->O26\n", "\n", "\n", "\n", "\n", "\n", "H8->O0\n", "\n", "\n", "\n", "\n", "\n", "H8->O1\n", "\n", "\n", "\n", "\n", "\n", "H8->O2\n", "\n", "\n", "\n", "\n", "\n", "H8->O3\n", "\n", "\n", "\n", "\n", "\n", "H8->O4\n", "\n", "\n", "\n", "\n", "\n", "H8->O5\n", "\n", "\n", "\n", "\n", "\n", "H8->O6\n", "\n", "\n", "\n", "\n", "\n", "H8->O7\n", "\n", "\n", "\n", "\n", "\n", "H8->O8\n", "\n", "\n", "\n", "\n", "\n", "H8->O9\n", "\n", "\n", "\n", "\n", "\n", "H8->O10\n", "\n", "\n", "\n", "\n", "\n", "H8->O11\n", "\n", "\n", "\n", "\n", "\n", "H8->O12\n", "\n", "\n", "\n", "\n", "\n", "H8->O13\n", "\n", "\n", "\n", "\n", "\n", "H8->O14\n", "\n", "\n", "\n", "\n", "\n", "H8->O15\n", "\n", "\n", "\n", "\n", "\n", "H8->O16\n", "\n", "\n", "\n", "\n", "\n", "H8->O17\n", "\n", "\n", "\n", "\n", "\n", "H8->O18\n", "\n", "\n", "\n", "\n", "\n", "H8->O19\n", "\n", "\n", "\n", "\n", "\n", "H8->O20\n", "\n", "\n", "\n", "\n", "\n", "H8->O21\n", "\n", "\n", "\n", "\n", "\n", "H8->O22\n", "\n", "\n", "\n", "\n", "\n", "H8->O23\n", "\n", "\n", "\n", "\n", "\n", "H8->O24\n", "\n", "\n", "\n", "\n", "\n", "H8->O25\n", "\n", "\n", "\n", "\n", "\n", "H8->O26\n", "\n", "\n", "\n", "\n", "\n", "H9->O0\n", "\n", "\n", "\n", "\n", "\n", "H9->O1\n", "\n", "\n", "\n", "\n", "\n", "H9->O2\n", "\n", "\n", "\n", "\n", "\n", "H9->O3\n", "\n", "\n", "\n", "\n", "\n", "H9->O4\n", "\n", "\n", "\n", "\n", "\n", "H9->O5\n", "\n", "\n", "\n", "\n", "\n", "H9->O6\n", "\n", "\n", "\n", "\n", "\n", "H9->O7\n", "\n", "\n", "\n", "\n", "\n", "H9->O8\n", "\n", "\n", "\n", "\n", "\n", "H9->O9\n", "\n", "\n", "\n", "\n", "\n", "H9->O10\n", "\n", "\n", "\n", "\n", "\n", "H9->O11\n", "\n", "\n", "\n", "\n", "\n", "H9->O12\n", "\n", "\n", "\n", "\n", "\n", "H9->O13\n", "\n", "\n", "\n", "\n", "\n", "H9->O14\n", "\n", "\n", "\n", "\n", "\n", "H9->O15\n", "\n", "\n", "\n", "\n", "\n", "H9->O16\n", "\n", "\n", "\n", "\n", "\n", "H9->O17\n", "\n", "\n", "\n", "\n", "\n", "H9->O18\n", "\n", "\n", "\n", "\n", "\n", "H9->O19\n", "\n", "\n", "\n", "\n", "\n", "H9->O20\n", "\n", "\n", "\n", "\n", "\n", "H9->O21\n", "\n", "\n", "\n", "\n", "\n", "H9->O22\n", "\n", "\n", "\n", "\n", "\n", "H9->O23\n", "\n", "\n", "\n", "\n", "\n", "H9->O24\n", "\n", "\n", "\n", "\n", "\n", "H9->O25\n", "\n", "\n", "\n", "\n", "\n", "H9->O26\n", "\n", "\n", "\n", "\n", "\n", "H10->O0\n", "\n", "\n", "\n", "\n", "\n", "H10->O1\n", "\n", "\n", "\n", "\n", "\n", "H10->O2\n", "\n", "\n", "\n", "\n", "\n", "H10->O3\n", "\n", "\n", "\n", "\n", "\n", "H10->O4\n", "\n", "\n", "\n", "\n", "\n", "H10->O5\n", "\n", "\n", "\n", "\n", "\n", "H10->O6\n", "\n", "\n", "\n", "\n", "\n", "H10->O7\n", "\n", "\n", "\n", "\n", "\n", "H10->O8\n", "\n", "\n", "\n", "\n", "\n", "H10->O9\n", "\n", "\n", "\n", "\n", "\n", "H10->O10\n", "\n", "\n", "\n", "\n", "\n", "H10->O11\n", "\n", "\n", "\n", "\n", "\n", "H10->O12\n", "\n", "\n", "\n", "\n", "\n", "H10->O13\n", "\n", "\n", "\n", "\n", "\n", "H10->O14\n", "\n", "\n", "\n", "\n", "\n", "H10->O15\n", "\n", "\n", "\n", "\n", "\n", "H10->O16\n", "\n", "\n", "\n", "\n", "\n", "H10->O17\n", "\n", "\n", "\n", "\n", "\n", "H10->O18\n", "\n", "\n", "\n", "\n", "\n", "H10->O19\n", "\n", "\n", "\n", "\n", "\n", "H10->O20\n", "\n", "\n", "\n", "\n", "\n", "H10->O21\n", "\n", "\n", "\n", "\n", "\n", "H10->O22\n", "\n", "\n", "\n", "\n", "\n", "H10->O23\n", "\n", "\n", "\n", "\n", "\n", "H10->O24\n", "\n", "\n", "\n", "\n", "\n", "H10->O25\n", "\n", "\n", "\n", "\n", "\n", "H10->O26\n", "\n", "\n", "\n", "\n", "\n", "H11->O0\n", "\n", "\n", "\n", "\n", "\n", "H11->O1\n", "\n", "\n", "\n", "\n", "\n", "H11->O2\n", "\n", "\n", "\n", "\n", "\n", "H11->O3\n", "\n", "\n", "\n", "\n", "\n", "H11->O4\n", "\n", "\n", "\n", "\n", "\n", "H11->O5\n", "\n", "\n", "\n", "\n", "\n", "H11->O6\n", "\n", "\n", "\n", "\n", "\n", "H11->O7\n", "\n", "\n", "\n", "\n", "\n", "H11->O8\n", "\n", "\n", "\n", "\n", "\n", "H11->O9\n", "\n", "\n", "\n", "\n", "\n", "H11->O10\n", "\n", "\n", "\n", "\n", "\n", "H11->O11\n", "\n", "\n", "\n", "\n", "\n", "H11->O12\n", "\n", "\n", "\n", "\n", "\n", "H11->O13\n", "\n", "\n", "\n", "\n", "\n", "H11->O14\n", "\n", "\n", "\n", "\n", "\n", "H11->O15\n", "\n", "\n", "\n", "\n", "\n", "H11->O16\n", "\n", "\n", "\n", "\n", "\n", "H11->O17\n", "\n", "\n", "\n", "\n", "\n", "H11->O18\n", "\n", "\n", "\n", "\n", "\n", "H11->O19\n", "\n", "\n", "\n", "\n", "\n", "H11->O20\n", "\n", "\n", "\n", "\n", "\n", "H11->O21\n", "\n", "\n", "\n", "\n", "\n", "H11->O22\n", "\n", "\n", "\n", "\n", "\n", "H11->O23\n", "\n", "\n", "\n", "\n", "\n", "H11->O24\n", "\n", "\n", "\n", "\n", "\n", "H11->O25\n", "\n", "\n", "\n", "\n", "\n", "H11->O26\n", "\n", "\n", "\n", "\n", "\n", "H12->O0\n", "\n", "\n", "\n", "\n", "\n", "H12->O1\n", "\n", "\n", "\n", "\n", "\n", "H12->O2\n", "\n", "\n", "\n", "\n", "\n", "H12->O3\n", "\n", "\n", "\n", "\n", "\n", "H12->O4\n", "\n", "\n", "\n", "\n", "\n", "H12->O5\n", "\n", "\n", "\n", "\n", "\n", "H12->O6\n", "\n", "\n", "\n", "\n", "\n", "H12->O7\n", "\n", "\n", "\n", "\n", "\n", "H12->O8\n", "\n", "\n", "\n", "\n", "\n", "H12->O9\n", "\n", "\n", "\n", "\n", "\n", "H12->O10\n", "\n", "\n", "\n", "\n", "\n", "H12->O11\n", "\n", "\n", "\n", "\n", "\n", "H12->O12\n", "\n", "\n", "\n", "\n", "\n", "H12->O13\n", "\n", "\n", "\n", "\n", "\n", "H12->O14\n", "\n", "\n", "\n", "\n", "\n", "H12->O15\n", "\n", "\n", "\n", "\n", "\n", "H12->O16\n", "\n", "\n", "\n", "\n", "\n", "H12->O17\n", "\n", "\n", "\n", "\n", "\n", "H12->O18\n", "\n", "\n", "\n", "\n", "\n", "H12->O19\n", "\n", "\n", "\n", "\n", "\n", "H12->O20\n", "\n", "\n", "\n", "\n", "\n", "H12->O21\n", "\n", "\n", "\n", "\n", "\n", "H12->O22\n", "\n", "\n", "\n", "\n", "\n", "H12->O23\n", "\n", "\n", "\n", "\n", "\n", "H12->O24\n", "\n", "\n", "\n", "\n", "\n", "H12->O25\n", "\n", "\n", "\n", "\n", "\n", "H12->O26\n", "\n", "\n", "\n", "\n", "\n", "H13->O0\n", "\n", "\n", "\n", "\n", "\n", "H13->O1\n", "\n", "\n", "\n", "\n", "\n", "H13->O2\n", "\n", "\n", "\n", "\n", "\n", "H13->O3\n", "\n", "\n", "\n", "\n", "\n", "H13->O4\n", "\n", "\n", "\n", "\n", "\n", "H13->O5\n", "\n", "\n", "\n", "\n", "\n", "H13->O6\n", "\n", "\n", "\n", "\n", "\n", "H13->O7\n", "\n", "\n", "\n", "\n", "\n", "H13->O8\n", "\n", "\n", "\n", "\n", "\n", "H13->O9\n", "\n", "\n", "\n", "\n", "\n", "H13->O10\n", "\n", "\n", "\n", "\n", "\n", "H13->O11\n", "\n", "\n", "\n", "\n", "\n", "H13->O12\n", "\n", "\n", "\n", "\n", "\n", "H13->O13\n", "\n", "\n", "\n", "\n", "\n", "H13->O14\n", "\n", "\n", "\n", "\n", "\n", "H13->O15\n", "\n", "\n", "\n", "\n", "\n", "H13->O16\n", "\n", "\n", "\n", "\n", "\n", "H13->O17\n", "\n", "\n", "\n", "\n", "\n", "H13->O18\n", "\n", "\n", "\n", "\n", "\n", "H13->O19\n", "\n", "\n", "\n", "\n", "\n", "H13->O20\n", "\n", "\n", "\n", "\n", "\n", "H13->O21\n", "\n", "\n", "\n", "\n", "\n", "H13->O22\n", "\n", "\n", "\n", "\n", "\n", "H13->O23\n", "\n", "\n", "\n", "\n", "\n", "H13->O24\n", "\n", "\n", "\n", "\n", "\n", "H13->O25\n", "\n", "\n", "\n", "\n", "\n", "H13->O26\n", "\n", "\n", "\n", "\n", "\n", "H14->O0\n", "\n", "\n", "\n", "\n", "\n", "H14->O1\n", "\n", "\n", "\n", "\n", "\n", "H14->O2\n", "\n", "\n", "\n", "\n", "\n", "H14->O3\n", "\n", "\n", "\n", "\n", "\n", "H14->O4\n", "\n", "\n", "\n", "\n", "\n", "H14->O5\n", "\n", "\n", "\n", "\n", "\n", "H14->O6\n", "\n", "\n", "\n", "\n", "\n", "H14->O7\n", "\n", "\n", "\n", "\n", "\n", "H14->O8\n", "\n", "\n", "\n", "\n", "\n", "H14->O9\n", "\n", "\n", "\n", "\n", "\n", "H14->O10\n", "\n", "\n", "\n", "\n", "\n", "H14->O11\n", "\n", "\n", "\n", "\n", "\n", "H14->O12\n", "\n", "\n", "\n", "\n", "\n", "H14->O13\n", "\n", "\n", "\n", "\n", "\n", "H14->O14\n", "\n", "\n", "\n", "\n", "\n", "H14->O15\n", "\n", "\n", "\n", "\n", "\n", "H14->O16\n", "\n", "\n", "\n", "\n", "\n", "H14->O17\n", "\n", "\n", "\n", "\n", "\n", "H14->O18\n", "\n", "\n", "\n", "\n", "\n", "H14->O19\n", "\n", "\n", "\n", "\n", "\n", "H14->O20\n", "\n", "\n", "\n", "\n", "\n", "H14->O21\n", "\n", "\n", "\n", "\n", "\n", "H14->O22\n", "\n", "\n", "\n", "\n", "\n", "H14->O23\n", "\n", "\n", "\n", "\n", "\n", "H14->O24\n", "\n", "\n", "\n", "\n", "\n", "H14->O25\n", "\n", "\n", "\n", "\n", "\n", "H14->O26\n", "\n", "\n", "\n", "\n", "\n", "H15->O0\n", "\n", "\n", "\n", "\n", "\n", "H15->O1\n", "\n", "\n", "\n", "\n", "\n", "H15->O2\n", "\n", "\n", "\n", "\n", "\n", "H15->O3\n", "\n", "\n", "\n", "\n", "\n", "H15->O4\n", "\n", "\n", "\n", "\n", "\n", "H15->O5\n", "\n", "\n", "\n", "\n", "\n", "H15->O6\n", "\n", "\n", "\n", "\n", "\n", "H15->O7\n", "\n", "\n", "\n", "\n", "\n", "H15->O8\n", "\n", "\n", "\n", "\n", "\n", "H15->O9\n", "\n", "\n", "\n", "\n", "\n", "H15->O10\n", "\n", "\n", "\n", "\n", "\n", "H15->O11\n", "\n", "\n", "\n", "\n", "\n", "H15->O12\n", "\n", "\n", "\n", "\n", "\n", "H15->O13\n", "\n", "\n", "\n", "\n", "\n", "H15->O14\n", "\n", "\n", "\n", "\n", "\n", "H15->O15\n", "\n", "\n", "\n", "\n", "\n", "H15->O16\n", "\n", "\n", "\n", "\n", "\n", "H15->O17\n", "\n", "\n", "\n", "\n", "\n", "H15->O18\n", "\n", "\n", "\n", "\n", "\n", "H15->O19\n", "\n", "\n", "\n", "\n", "\n", "H15->O20\n", "\n", "\n", "\n", "\n", "\n", "H15->O21\n", "\n", "\n", "\n", "\n", "\n", "H15->O22\n", "\n", "\n", "\n", "\n", "\n", "H15->O23\n", "\n", "\n", "\n", "\n", "\n", "H15->O24\n", "\n", "\n", "\n", "\n", "\n", "H15->O25\n", "\n", "\n", "\n", "\n", "\n", "H15->O26\n", "\n", "\n", "\n", "\n", "\n", "H16->O0\n", "\n", "\n", "\n", "\n", "\n", "H16->O1\n", "\n", "\n", "\n", "\n", "\n", "H16->O2\n", "\n", "\n", "\n", "\n", "\n", "H16->O3\n", "\n", "\n", "\n", "\n", "\n", "H16->O4\n", "\n", "\n", "\n", "\n", "\n", "H16->O5\n", "\n", "\n", "\n", "\n", "\n", "H16->O6\n", "\n", "\n", "\n", "\n", "\n", "H16->O7\n", "\n", "\n", "\n", "\n", "\n", "H16->O8\n", "\n", "\n", "\n", "\n", "\n", "H16->O9\n", "\n", "\n", "\n", "\n", "\n", "H16->O10\n", "\n", "\n", "\n", "\n", "\n", "H16->O11\n", "\n", "\n", "\n", "\n", "\n", "H16->O12\n", "\n", "\n", "\n", "\n", "\n", "H16->O13\n", "\n", "\n", "\n", "\n", "\n", "H16->O14\n", "\n", "\n", "\n", "\n", "\n", "H16->O15\n", "\n", "\n", "\n", "\n", "\n", "H16->O16\n", "\n", "\n", "\n", "\n", "\n", "H16->O17\n", "\n", "\n", "\n", "\n", "\n", "H16->O18\n", "\n", "\n", "\n", "\n", "\n", "H16->O19\n", "\n", "\n", "\n", "\n", "\n", "H16->O20\n", "\n", "\n", "\n", "\n", "\n", "H16->O21\n", "\n", "\n", "\n", "\n", "\n", "H16->O22\n", "\n", "\n", "\n", "\n", "\n", "H16->O23\n", "\n", "\n", "\n", "\n", "\n", "H16->O24\n", "\n", "\n", "\n", "\n", "\n", "H16->O25\n", "\n", "\n", "\n", "\n", "\n", "H16->O26\n", "\n", "\n", "\n", "\n", "\n", "H17->O0\n", "\n", "\n", "\n", "\n", "\n", "H17->O1\n", "\n", "\n", "\n", "\n", "\n", "H17->O2\n", "\n", "\n", "\n", "\n", "\n", "H17->O3\n", "\n", "\n", "\n", "\n", "\n", "H17->O4\n", "\n", "\n", "\n", "\n", "\n", "H17->O5\n", "\n", "\n", "\n", "\n", "\n", "H17->O6\n", "\n", "\n", "\n", "\n", "\n", "H17->O7\n", "\n", "\n", "\n", "\n", "\n", "H17->O8\n", "\n", "\n", "\n", "\n", "\n", "H17->O9\n", "\n", "\n", "\n", "\n", "\n", "H17->O10\n", "\n", "\n", "\n", "\n", "\n", "H17->O11\n", "\n", "\n", "\n", "\n", "\n", "H17->O12\n", "\n", "\n", "\n", "\n", "\n", "H17->O13\n", "\n", "\n", "\n", "\n", "\n", "H17->O14\n", "\n", "\n", "\n", "\n", "\n", "H17->O15\n", "\n", "\n", "\n", "\n", "\n", "H17->O16\n", "\n", "\n", "\n", "\n", "\n", "H17->O17\n", "\n", "\n", "\n", "\n", "\n", "H17->O18\n", "\n", "\n", "\n", "\n", "\n", "H17->O19\n", "\n", "\n", "\n", "\n", "\n", "H17->O20\n", "\n", "\n", "\n", "\n", "\n", "H17->O21\n", "\n", "\n", "\n", "\n", "\n", "H17->O22\n", "\n", "\n", "\n", "\n", "\n", "H17->O23\n", "\n", "\n", "\n", "\n", "\n", "H17->O24\n", "\n", "\n", "\n", "\n", "\n", "H17->O25\n", "\n", "\n", "\n", "\n", "\n", "H17->O26\n", "\n", "\n", "\n", "\n", "\n", "H18->O0\n", "\n", "\n", "\n", "\n", "\n", "H18->O1\n", "\n", "\n", "\n", "\n", "\n", "H18->O2\n", "\n", "\n", "\n", "\n", "\n", "H18->O3\n", "\n", "\n", "\n", "\n", "\n", "H18->O4\n", "\n", "\n", "\n", "\n", "\n", "H18->O5\n", "\n", "\n", "\n", "\n", "\n", "H18->O6\n", "\n", "\n", "\n", "\n", "\n", "H18->O7\n", "\n", "\n", "\n", "\n", "\n", "H18->O8\n", "\n", "\n", "\n", "\n", "\n", "H18->O9\n", "\n", "\n", "\n", "\n", "\n", "H18->O10\n", "\n", "\n", "\n", "\n", "\n", "H18->O11\n", "\n", "\n", "\n", "\n", "\n", "H18->O12\n", "\n", "\n", "\n", "\n", "\n", "H18->O13\n", "\n", "\n", "\n", "\n", "\n", "H18->O14\n", "\n", "\n", "\n", "\n", "\n", "H18->O15\n", "\n", "\n", "\n", "\n", "\n", "H18->O16\n", "\n", "\n", "\n", "\n", "\n", "H18->O17\n", "\n", "\n", "\n", "\n", "\n", "H18->O18\n", "\n", "\n", "\n", "\n", "\n", "H18->O19\n", "\n", "\n", "\n", "\n", "\n", "H18->O20\n", "\n", "\n", "\n", "\n", "\n", "H18->O21\n", "\n", "\n", "\n", "\n", "\n", "H18->O22\n", "\n", "\n", "\n", "\n", "\n", "H18->O23\n", "\n", "\n", "\n", "\n", "\n", "H18->O24\n", "\n", "\n", "\n", "\n", "\n", "H18->O25\n", "\n", "\n", "\n", "\n", "\n", "H18->O26\n", "\n", "\n", "\n", "\n", "\n", "H19->O0\n", "\n", "\n", "\n", "\n", "\n", "H19->O1\n", "\n", "\n", "\n", "\n", "\n", "H19->O2\n", "\n", "\n", "\n", "\n", "\n", "H19->O3\n", "\n", "\n", "\n", "\n", "\n", "H19->O4\n", "\n", "\n", "\n", "\n", "\n", "H19->O5\n", "\n", "\n", "\n", "\n", "\n", "H19->O6\n", "\n", "\n", "\n", "\n", "\n", "H19->O7\n", "\n", "\n", "\n", "\n", "\n", "H19->O8\n", "\n", "\n", "\n", "\n", "\n", "H19->O9\n", "\n", "\n", "\n", "\n", "\n", "H19->O10\n", "\n", "\n", "\n", "\n", "\n", "H19->O11\n", "\n", "\n", "\n", "\n", "\n", "H19->O12\n", "\n", "\n", "\n", "\n", "\n", "H19->O13\n", "\n", "\n", "\n", "\n", "\n", "H19->O14\n", "\n", "\n", "\n", "\n", "\n", "H19->O15\n", "\n", "\n", "\n", "\n", "\n", "H19->O16\n", "\n", "\n", "\n", "\n", "\n", "H19->O17\n", "\n", "\n", "\n", "\n", "\n", "H19->O18\n", "\n", "\n", "\n", "\n", "\n", "H19->O19\n", "\n", "\n", "\n", "\n", "\n", "H19->O20\n", "\n", "\n", "\n", "\n", "\n", "H19->O21\n", "\n", "\n", "\n", "\n", "\n", "H19->O22\n", "\n", "\n", "\n", "\n", "\n", "H19->O23\n", "\n", "\n", "\n", "\n", "\n", "H19->O24\n", "\n", "\n", "\n", "\n", "\n", "H19->O25\n", "\n", "\n", "\n", "\n", "\n", "H19->O26\n", "\n", "\n", "\n", "\n", "\n", "H20->O0\n", "\n", "\n", "\n", "\n", "\n", "H20->O1\n", "\n", "\n", "\n", "\n", "\n", "H20->O2\n", "\n", "\n", "\n", "\n", "\n", "H20->O3\n", "\n", "\n", "\n", "\n", "\n", "H20->O4\n", "\n", "\n", "\n", "\n", "\n", "H20->O5\n", "\n", "\n", "\n", "\n", "\n", "H20->O6\n", "\n", "\n", "\n", "\n", "\n", "H20->O7\n", "\n", "\n", "\n", "\n", "\n", "H20->O8\n", "\n", "\n", "\n", "\n", "\n", "H20->O9\n", "\n", "\n", "\n", "\n", "\n", "H20->O10\n", "\n", "\n", "\n", "\n", "\n", "H20->O11\n", "\n", "\n", "\n", "\n", "\n", "H20->O12\n", "\n", "\n", "\n", "\n", "\n", "H20->O13\n", "\n", "\n", "\n", "\n", "\n", "H20->O14\n", "\n", "\n", "\n", "\n", "\n", "H20->O15\n", "\n", "\n", "\n", "\n", "\n", "H20->O16\n", "\n", "\n", "\n", "\n", "\n", "H20->O17\n", "\n", "\n", "\n", "\n", "\n", "H20->O18\n", "\n", "\n", "\n", "\n", "\n", "H20->O19\n", "\n", "\n", "\n", "\n", "\n", "H20->O20\n", "\n", "\n", "\n", "\n", "\n", "H20->O21\n", "\n", "\n", "\n", "\n", "\n", "H20->O22\n", "\n", "\n", "\n", "\n", "\n", "H20->O23\n", "\n", "\n", "\n", "\n", "\n", "H20->O24\n", "\n", "\n", "\n", "\n", "\n", "H20->O25\n", "\n", "\n", "\n", "\n", "\n", "H20->O26\n", "\n", "\n", "\n", "\n", "\n", "H21->O0\n", "\n", "\n", "\n", "\n", "\n", "H21->O1\n", "\n", "\n", "\n", "\n", "\n", "H21->O2\n", "\n", "\n", "\n", "\n", "\n", "H21->O3\n", "\n", "\n", "\n", "\n", "\n", "H21->O4\n", "\n", "\n", "\n", "\n", "\n", "H21->O5\n", "\n", "\n", "\n", "\n", "\n", "H21->O6\n", "\n", "\n", "\n", "\n", "\n", "H21->O7\n", "\n", "\n", "\n", "\n", "\n", "H21->O8\n", "\n", "\n", "\n", "\n", "\n", "H21->O9\n", "\n", "\n", "\n", "\n", "\n", "H21->O10\n", "\n", "\n", "\n", "\n", "\n", "H21->O11\n", "\n", "\n", "\n", "\n", "\n", "H21->O12\n", "\n", "\n", "\n", "\n", "\n", "H21->O13\n", "\n", "\n", "\n", "\n", "\n", "H21->O14\n", "\n", "\n", "\n", "\n", "\n", "H21->O15\n", "\n", "\n", "\n", "\n", "\n", "H21->O16\n", "\n", "\n", "\n", "\n", "\n", "H21->O17\n", "\n", "\n", "\n", "\n", "\n", "H21->O18\n", "\n", "\n", "\n", "\n", "\n", "H21->O19\n", "\n", "\n", "\n", "\n", "\n", "H21->O20\n", "\n", "\n", "\n", "\n", "\n", "H21->O21\n", "\n", "\n", "\n", "\n", "\n", "H21->O22\n", "\n", "\n", "\n", "\n", "\n", "H21->O23\n", "\n", "\n", "\n", "\n", "\n", "H21->O24\n", "\n", "\n", "\n", "\n", "\n", "H21->O25\n", "\n", "\n", "\n", "\n", "\n", "H21->O26\n", "\n", "\n", "\n", "\n", "\n", "H22->O0\n", "\n", "\n", "\n", "\n", "\n", "H22->O1\n", "\n", "\n", "\n", "\n", "\n", "H22->O2\n", "\n", "\n", "\n", "\n", "\n", "H22->O3\n", "\n", "\n", "\n", "\n", "\n", "H22->O4\n", "\n", "\n", "\n", "\n", "\n", "H22->O5\n", "\n", "\n", "\n", "\n", "\n", "H22->O6\n", "\n", "\n", "\n", "\n", "\n", "H22->O7\n", "\n", "\n", "\n", "\n", "\n", "H22->O8\n", "\n", "\n", "\n", "\n", "\n", "H22->O9\n", "\n", "\n", "\n", "\n", "\n", "H22->O10\n", "\n", "\n", "\n", "\n", "\n", "H22->O11\n", "\n", "\n", "\n", "\n", "\n", "H22->O12\n", "\n", "\n", "\n", "\n", "\n", "H22->O13\n", "\n", "\n", "\n", "\n", "\n", "H22->O14\n", "\n", "\n", "\n", "\n", "\n", "H22->O15\n", "\n", "\n", "\n", "\n", "\n", "H22->O16\n", "\n", "\n", "\n", "\n", "\n", "H22->O17\n", "\n", "\n", "\n", "\n", "\n", "H22->O18\n", "\n", "\n", "\n", "\n", "\n", "H22->O19\n", "\n", "\n", "\n", "\n", "\n", "H22->O20\n", "\n", "\n", "\n", "\n", "\n", "H22->O21\n", "\n", "\n", "\n", "\n", "\n", "H22->O22\n", "\n", "\n", "\n", "\n", "\n", "H22->O23\n", "\n", "\n", "\n", "\n", "\n", "H22->O24\n", "\n", "\n", "\n", "\n", "\n", "H22->O25\n", "\n", "\n", "\n", "\n", "\n", "H22->O26\n", "\n", "\n", "\n", "\n", "\n", "H23->O0\n", "\n", "\n", "\n", "\n", "\n", "H23->O1\n", "\n", "\n", "\n", "\n", "\n", "H23->O2\n", "\n", "\n", "\n", "\n", "\n", "H23->O3\n", "\n", "\n", "\n", "\n", "\n", "H23->O4\n", "\n", "\n", "\n", "\n", "\n", "H23->O5\n", "\n", "\n", "\n", "\n", "\n", "H23->O6\n", "\n", "\n", "\n", "\n", "\n", "H23->O7\n", "\n", "\n", "\n", "\n", "\n", "H23->O8\n", "\n", "\n", "\n", "\n", "\n", "H23->O9\n", "\n", "\n", "\n", "\n", "\n", "H23->O10\n", "\n", "\n", "\n", "\n", "\n", "H23->O11\n", "\n", "\n", "\n", "\n", "\n", "H23->O12\n", "\n", "\n", "\n", "\n", "\n", "H23->O13\n", "\n", "\n", "\n", "\n", "\n", "H23->O14\n", "\n", "\n", "\n", "\n", "\n", "H23->O15\n", "\n", "\n", "\n", "\n", "\n", "H23->O16\n", "\n", "\n", "\n", "\n", "\n", "H23->O17\n", "\n", "\n", "\n", "\n", "\n", "H23->O18\n", "\n", "\n", "\n", "\n", "\n", "H23->O19\n", "\n", "\n", "\n", "\n", "\n", "H23->O20\n", "\n", "\n", "\n", "\n", "\n", "H23->O21\n", "\n", "\n", "\n", "\n", "\n", "H23->O22\n", "\n", "\n", "\n", "\n", "\n", "H23->O23\n", "\n", "\n", "\n", "\n", "\n", "H23->O24\n", "\n", "\n", "\n", "\n", "\n", "H23->O25\n", "\n", "\n", "\n", "\n", "\n", "H23->O26\n", "\n", "\n", "\n", "\n", "\n", "H24->O0\n", "\n", "\n", "\n", "\n", "\n", "H24->O1\n", "\n", "\n", "\n", "\n", "\n", "H24->O2\n", "\n", "\n", "\n", "\n", "\n", "H24->O3\n", "\n", "\n", "\n", "\n", "\n", "H24->O4\n", "\n", "\n", "\n", "\n", "\n", "H24->O5\n", "\n", "\n", "\n", "\n", "\n", "H24->O6\n", "\n", "\n", "\n", "\n", "\n", "H24->O7\n", "\n", "\n", "\n", "\n", "\n", "H24->O8\n", "\n", "\n", "\n", "\n", "\n", "H24->O9\n", "\n", "\n", "\n", "\n", "\n", "H24->O10\n", "\n", "\n", "\n", "\n", "\n", "H24->O11\n", "\n", "\n", "\n", "\n", "\n", "H24->O12\n", "\n", "\n", "\n", "\n", "\n", "H24->O13\n", "\n", "\n", "\n", "\n", "\n", "H24->O14\n", "\n", "\n", "\n", "\n", "\n", "H24->O15\n", "\n", "\n", "\n", "\n", "\n", "H24->O16\n", "\n", "\n", "\n", "\n", "\n", "H24->O17\n", "\n", "\n", "\n", "\n", "\n", "H24->O18\n", "\n", "\n", "\n", "\n", "\n", "H24->O19\n", "\n", "\n", "\n", "\n", "\n", "H24->O20\n", "\n", "\n", "\n", "\n", "\n", "H24->O21\n", "\n", "\n", "\n", "\n", "\n", "H24->O22\n", "\n", "\n", "\n", "\n", "\n", "H24->O23\n", "\n", "\n", "\n", "\n", "\n", "H24->O24\n", "\n", "\n", "\n", "\n", "\n", "H24->O25\n", "\n", "\n", "\n", "\n", "\n", "H24->O26\n", "\n", "\n", "\n", "\n", "\n", "H25->O0\n", "\n", "\n", "\n", "\n", "\n", "H25->O1\n", "\n", "\n", "\n", "\n", "\n", "H25->O2\n", "\n", "\n", "\n", "\n", "\n", "H25->O3\n", "\n", "\n", "\n", "\n", "\n", "H25->O4\n", "\n", "\n", "\n", "\n", "\n", "H25->O5\n", "\n", "\n", "\n", "\n", "\n", "H25->O6\n", "\n", "\n", "\n", "\n", "\n", "H25->O7\n", "\n", "\n", "\n", "\n", "\n", "H25->O8\n", "\n", "\n", "\n", "\n", "\n", "H25->O9\n", "\n", "\n", "\n", "\n", "\n", "H25->O10\n", "\n", "\n", "\n", "\n", "\n", "H25->O11\n", "\n", "\n", "\n", "\n", "\n", "H25->O12\n", "\n", "\n", "\n", "\n", "\n", "H25->O13\n", "\n", "\n", "\n", "\n", "\n", "H25->O14\n", "\n", "\n", "\n", "\n", "\n", "H25->O15\n", "\n", "\n", "\n", "\n", "\n", "H25->O16\n", "\n", "\n", "\n", "\n", "\n", "H25->O17\n", "\n", "\n", "\n", "\n", "\n", "H25->O18\n", "\n", "\n", "\n", "\n", "\n", "H25->O19\n", "\n", "\n", "\n", "\n", "\n", "H25->O20\n", "\n", "\n", "\n", "\n", "\n", "H25->O21\n", "\n", "\n", "\n", "\n", "\n", "H25->O22\n", "\n", "\n", "\n", "\n", "\n", "H25->O23\n", "\n", "\n", "\n", "\n", "\n", "H25->O24\n", "\n", "\n", "\n", "\n", "\n", "H25->O25\n", "\n", "\n", "\n", "\n", "\n", "H25->O26\n", "\n", "\n", "\n", "\n", "\n", "H26->O0\n", "\n", "\n", "\n", "\n", "\n", "H26->O1\n", "\n", "\n", "\n", "\n", "\n", "H26->O2\n", "\n", "\n", "\n", "\n", "\n", "H26->O3\n", "\n", "\n", "\n", "\n", "\n", "H26->O4\n", "\n", "\n", "\n", "\n", "\n", "H26->O5\n", "\n", "\n", "\n", "\n", "\n", "H26->O6\n", "\n", "\n", "\n", "\n", "\n", "H26->O7\n", "\n", "\n", "\n", "\n", "\n", "H26->O8\n", "\n", "\n", "\n", "\n", "\n", "H26->O9\n", "\n", "\n", "\n", "\n", "\n", "H26->O10\n", "\n", "\n", "\n", "\n", "\n", "H26->O11\n", "\n", "\n", "\n", "\n", "\n", "H26->O12\n", "\n", "\n", "\n", "\n", "\n", "H26->O13\n", "\n", "\n", "\n", "\n", "\n", "H26->O14\n", "\n", "\n", "\n", "\n", "\n", "H26->O15\n", "\n", "\n", "\n", "\n", "\n", "H26->O16\n", "\n", "\n", "\n", "\n", "\n", "H26->O17\n", "\n", "\n", "\n", "\n", "\n", "H26->O18\n", "\n", "\n", "\n", "\n", "\n", "H26->O19\n", "\n", "\n", "\n", "\n", "\n", "H26->O20\n", "\n", "\n", "\n", "\n", "\n", "H26->O21\n", "\n", "\n", "\n", "\n", "\n", "H26->O22\n", "\n", "\n", "\n", "\n", "\n", "H26->O23\n", "\n", "\n", "\n", "\n", "\n", "H26->O24\n", "\n", "\n", "\n", "\n", "\n", "H26->O25\n", "\n", "\n", "\n", "\n", "\n", "H26->O26\n", "\n", "\n", "\n", "\n", "\n" ], "text/plain": [ "" ] }, "execution_count": 11, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Let us visualize the neural network we will use for this task\n", "\n", "from graphviz import Digraph\n", "\n", "# Create a new directed graph\n", "dot = Digraph()\n", "\n", "# Set graph attributes for layout\n", "dot.attr(rankdir='LR') # Left to right layout\n", "dot.attr('graph', nodesep='1', ranksep='2') # Increase node separation for width\n", "\n", "# Set node attributes\n", "dot.attr('node', shape='circle') # Use circle for neurons\n", "\n", "# Add input layer nodes (27 inputs)\n", "for i in range(27):\n", " dot.node(f'I{i}', label=f'Input {i+1}')\n", "\n", "# Add hidden layer nodes (27 neurons)\n", "for h in range(27):\n", " dot.node(f'H{h}', label=f'H {h+1}')\n", "\n", "# Add output layer nodes (27 output neurons with softmax)\n", "for o in range(27):\n", " dot.node(f'O{o}', label=f'Softmax {o+1}')\n", "\n", "# Add edges from input layer to hidden layer (fully connected)\n", "for i in range(27):\n", " for h in range(27):\n", " dot.edge(f'I{i}', f'H{h}')\n", "\n", "# Add edges from hidden layer to output layer (fully connected to each output node)\n", "for h in range(27):\n", " for o in range(27):\n", " dot.edge(f'H{h}', f'O{o}')\n", "\n", "# Render the graph in a file (e.g., as PNG)\n", "dot.render('neural_network_wide', format='png', cleanup=False)\n", "\n", "# Optionally, display it directly in Jupyter if running in a notebook\n", "dot.render('neural_network_wide', format='png')\n", "dot" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [], "source": [ "# Create a \"naive\" neural network with 27x27 weights -> 729 parameters to learn\n", "\n", "W = torch.randn((27, 27), requires_grad=True)" ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [], "source": [ "# Forward pass (one letter as input, one letter as output) with five training examples\n", "\n", "H = xs_onehot @ W\n", "output = torch.exp(H) / torch.exp(H).sum(1, keepdim=True) # This is the softmax function" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "tensor([0.0331, 0.0158, 0.0111, 0.0107, 0.0159, 0.0989, 0.0293, 0.0967, 0.0648,\n", " 0.0341, 0.0143, 0.0132, 0.0057, 0.0691, 0.0076, 0.0093, 0.0216, 0.0447,\n", " 0.0044, 0.0063, 0.0328, 0.0417, 0.2168, 0.0459, 0.0458, 0.0079, 0.0025],\n", " grad_fn=)\n", "tensor(1., grad_fn=)\n" ] } ], "source": [ "# The output can be interpreted as probabilities, here we print the probs for '.' (the first letter in the first word)\n", "\n", "print(output[0])\n", "print(sum(output[0]))" ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Plot output[0]\n", "plt.figure(figsize=(10, 5))\n", "plt.bar(range(27), output[0].detach().numpy())\n", "plt.xticks(range(27), map_i_to_s.values())\n", "plt.title('Probability that a letter is the first letter in a name (output[0])')\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor(0.0989, grad_fn=)" ] }, "execution_count": 16, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# What is the prediction (probability of the true value) for the first letter ('e')?\n", "output[0, 5]" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(tensor(0.0220, grad_fn=),\n", " tensor(0.0711, grad_fn=),\n", " tensor(0.0229, grad_fn=),\n", " tensor(0.0810, grad_fn=))" ] }, "execution_count": 17, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# What about the four next letters\n", "output[1, 13], output[2, 13], output[3, 1], output[4, 0]" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor([0.0989, 0.0220, 0.0711, 0.0229, 0.0810], grad_fn=)" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Lets generalize a bit (number of examples passed through the network mapped to the true y value)\n", "output[torch.arange(len(ys)), ys]" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor(3.0128, grad_fn=)" ] }, "execution_count": 19, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Calculate the loss\n", "\n", "loss = -output[torch.arange(len(ys)), ys].log().mean()\n", "loss" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [], "source": [ "# Backward pass\n", "# --------\n", "# Backpropagation is done by calling the backward() method on the loss tensor\n", "W.grad = None # set the gradient to zero\n", "loss.backward()" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [], "source": [ "# Let us update the weights\n", "\n", "# Learning rate\n", "lr = 1\n", "\n", "W.data -= lr * W.grad" ] }, { "cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [], "source": [ "# Another forward pass\n", "H = xs_onehot @ W\n", "output = torch.exp(H) / torch.exp(H).sum(1, keepdim=True)" ] }, { "cell_type": "code", "execution_count": 23, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tensor(2.8253, grad_fn=)" ] }, "execution_count": 23, "metadata": {}, "output_type": "execute_result" } ], "source": [ "loss = -output[torch.arange(len(ys)), ys].log().mean()\n", "loss" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Let us use all the data and train for many epochs" ] }, { "cell_type": "code", "execution_count": 24, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "torch.Size([228146, 27])" ] }, "execution_count": 24, "metadata": {}, "output_type": "execute_result" } ], "source": [ "## Let us automate the training process and run on all the names\n", "# Create the training set\n", "xs, ys = [], []\n", "\n", "for w in words:\n", " chs = ['.'] + list(w) + ['.']\n", " for ch1, ch2 in zip(chs, chs[1:]):\n", " ix1 = map_s_to_i[ch1]\n", " ix2 = map_s_to_i[ch2]\n", " xs.append(ix1)\n", " ys.append(ix2)\n", "\n", "xs = torch.tensor(xs)\n", "ys = torch.tensor(ys)\n", "\n", "xs_onehot = F.one_hot(xs, num_classes=27).float()\n", "xs_onehot.shape" ] }, { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [], "source": [ "# Create the neural network\n", "W = torch.randn((27, 27), requires_grad=True)" ] }, { "cell_type": "code", "execution_count": 26, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Loss: 3.9603922367095947\n", "Loss: 3.950573205947876\n", "Loss: 3.9408535957336426\n", "Loss: 3.9312307834625244\n", "Loss: 3.921703577041626\n", "Loss: 3.9122705459594727\n", "Loss: 3.9029297828674316\n", "Loss: 3.8936798572540283\n", "Loss: 3.884518623352051\n", "Loss: 3.875445604324341\n" ] } ], "source": [ "# Train the neural network for 10 iterations\n", "\n", "for i in range(10):\n", " H = xs_onehot @ W\n", " output = torch.exp(H) / torch.exp(H).sum(1, keepdim=True)\n", " loss = -output[torch.arange(len(ys)), ys].log().mean()\n", " W.grad = None\n", " loss.backward()\n", " W.data -= 1.0 * W.grad\n", " print(f'Loss: {loss.item()}')" ] }, { "cell_type": "code", "execution_count": 27, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Loss at iteration 0: 3.866459369659424\n", "Loss at iteration 100: 2.5297720432281494\n", "Loss at iteration 200: 2.4888861179351807\n", "Loss at iteration 300: 2.4758613109588623\n", "Loss at iteration 400: 2.469475746154785\n", "Loss at iteration 500: 2.4658145904541016\n", "Loss at iteration 600: 2.463522434234619\n", "Loss at iteration 700: 2.4619650840759277\n", "Loss at iteration 800: 2.460834264755249\n", "Loss at iteration 900: 2.4599735736846924\n" ] } ], "source": [ "# Train the neural network for 1_000 iterations with a higher learning rate\n", "\n", "for i in range(1_000):\n", " H = xs_onehot @ W\n", " output = torch.exp(H) / torch.exp(H).sum(1, keepdim=True)\n", " loss = -output[torch.arange(len(ys)), ys].log().mean()\n", " W.grad = None\n", " loss.backward()\n", " W.data -= 15.0 * W.grad\n", " if i % 100 == 0:\n", " print(f'Loss at iteration {i}: {loss.item()}')" ] }, { "cell_type": "code", "execution_count": 28, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Loss: 2.4592957496643066\n", "Loss: 2.4592955112457275\n", "Loss: 2.4592952728271484\n", "Loss: 2.4592947959899902\n", "Loss: 2.459294319152832\n", "Loss: 2.459294080734253\n", "Loss: 2.4592936038970947\n", "Loss: 2.4592931270599365\n", "Loss: 2.4592928886413574\n", "Loss: 2.45929217338562\n" ] } ], "source": [ "# Train the neural network for 10 iterations with a lower learning rate\n", "\n", "for i in range(10):\n", " H = xs_onehot @ W\n", " output = torch.exp(H) / torch.exp(H).sum(1, keepdim=True)\n", " loss = -output[torch.arange(len(ys)), ys].log().mean()\n", " W.grad = None\n", " loss.backward()\n", " W.data -= 1.0 * W.grad\n", " print(f'Loss: {loss.item()}')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# How can we generate names from our neural network?" ] }, { "cell_type": "code", "execution_count": 29, "metadata": {}, "outputs": [ { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Plot output[0]\n", "plt.figure(figsize=(10, 5))\n", "plt.bar(range(27), output[0].detach().numpy())\n", "plt.xticks(range(27), map_i_to_s.values())\n", "plt.title('Probability that a letter is the first letter in a name (output[0])')\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": 30, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "jarope\n", "krand\n", "kn\n", "ke\n", "di\n", "crlirreeia\n", "sorih\n", "mon\n", "garyre\n", "jariyovelee\n" ] } ], "source": [ "# Let us generate names from the neural net\n", "\n", "for i in range(10):\n", " out = []\n", " ix = 0\n", " while True:\n", " x_onehot = F.one_hot(torch.tensor([ix]), num_classes=27).float()\n", " H = x_onehot @ W\n", " output = torch.exp(H) / torch.exp(H).sum()\n", " ix = torch.multinomial(output, num_samples=1, replacement=True).item()\n", " out.append(map_i_to_s[ix])\n", " if len(out) > 15:\n", " break\n", " if ix == 0:\n", " if len(out) < 3:\n", " out = []\n", " ix = 0\n", " x_onehot = F.one_hot(torch.tensor([ix]), num_classes=27).float()\n", " continue\n", " break\n", " print(''.join(out)[0:-1])" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "pytorch", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.11.10" } }, "nbformat": 4, "nbformat_minor": 4 }