llm-from-scratch.ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":9197649,"sourceType":"datasetVersion","datasetId":5560635},{"sourceId":9287089,"sourceType":"datasetVersion","datasetId":5621924},{"sourceId":9367410,"sourceType":"datasetVersion","datasetId":5680682},{"sourceId":111833,"sourceType":"modelInstanceVersion","isSourceIdPinned":true,"modelInstanceId":93729,"modelId":117948}],"dockerImageVersionId":30746,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# LLM from scratch\n\nThis is a university diploma project crafted by Mihajlo Madzarevic with the help of lectures provided by Andrej Karpathy on making a LLM from scratch.","metadata":{}},{"cell_type":"markdown","source":"## Imports and data loading","metadata":{}},{"cell_type":"code","source":"# Module imports\n\nimport numpy as np\nimport pandas as pd\nimport matplotlib\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport torch\n# (nn - neural net) Module, dataset loading\nimport torch.nn as nn\n# conv, pool layers, attention mechanism, activation functions...\nfrom torch.nn import functional as F\n# Check whether we're on gpu or not\ndevice = 'cuda' if torch.cuda.is_available() else 'cpu'\nprint(\"Device used for training: %s\" % device)\n\nimport os\n\nprint(\"Loading the following texts: \")\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-09-30T13:04:39.391322Z","iopub.execute_input":"2024-09-30T13:04:39.392401Z","iopub.status.idle":"2024-09-30T13:04:44.836233Z","shell.execute_reply.started":"2024-09-30T13:04:39.392364Z","shell.execute_reply":"2024-09-30T13:04:44.835278Z"},"trusted":true},"execution_count":1,"outputs":[{"name":"stdout","text":"Device used for training: cuda\nLoading the following texts: \n/kaggle/input/meditations-marcus-aurelius/meditations.txt\n/kaggle/input/gpt-words/model_words_weights.pth\n/kaggle/input/llm-gpt/pytorch/default/1/model_words_weights BEST.pth\n/kaggle/input/transformer-weights/model_chars_weights.pth\n","output_type":"stream"}]},{"cell_type":"code","source":"# Params\n\n# These are more optimal for GPT model\n\n# How many tokens you want for predicting.\n# Tokens can be words/chars/subwords...\nblock_size = 256 #256\n# How many examples of blocks per batch for training.\nbatch_size = 64 #64\n# How many times the model should be evaluated.\neval_iters = 10\n# Number of embedding dimensions.\nn_embd = 384\n# Number of heads for a single Head of the GPT model\nn_heads = 6\n# Number of block layers in the GPT model.\nn_layers = 6\n# Dropout rate of the model.\ndropout = 0.2","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:04:57.347995Z","iopub.execute_input":"2024-09-30T13:04:57.348745Z","iopub.status.idle":"2024-09-30T13:04:57.353885Z","shell.execute_reply.started":"2024-09-30T13:04:57.348714Z","shell.execute_reply":"2024-09-30T13:04:57.352822Z"},"trusted":true},"execution_count":2,"outputs":[]},{"cell_type":"code","source":"# Symbols to remove from words.\nunwanted_symbols = \";!,./\\?&')(_\"\n# Optionally call before strip function with unwanted_symbols as argument\n# .replace('\"', '')\n# .replace('\"', '').strip(unwanted_symbols)","metadata":{"execution":{"iopub.status.busy":"2024-09-11T13:01:55.042590Z","iopub.execute_input":"2024-09-11T13:01:55.042955Z","iopub.status.idle":"2024-09-11T13:01:55.047137Z","shell.execute_reply.started":"2024-09-11T13:01:55.042928Z","shell.execute_reply":"2024-09-11T13:01:55.046191Z"},"trusted":true},"execution_count":4,"outputs":[]},{"cell_type":"code","source":"# Function that loads and cleans data.\ndef load_clean_data():\n    # Using with so it properly closes the resource\n    # in this case the file (we don't have to call file.close())\n    with open('/kaggle/input/meditations-marcus-aurelius/meditations.txt', 'r', encoding='utf-8') as file:\n        text = file.read()\n        \n        # CHAR PROCESSING\n        \n        chars_from_text = sorted(list(set(text.lower())))\n        \n        # Get rid of numbers and unnecessary symbols.\n        chars_from_text_tmp = []\n        for char in chars_from_text:\n            # Look for only letters by ASCII value.\n            if ord(char) >= 97 and ord(char) <= 122:\n                chars_from_text_tmp.append(char)\n                \n        # Manually add needed symbols.\n        chars_from_text_tmp.insert(0, \"?\")\n        chars_from_text_tmp.insert(0, \"!\")\n        chars_from_text_tmp.insert(0, \".\")\n        chars_from_text_tmp.insert(0, \" \")\n        \n        chars_from_text = chars_from_text_tmp\n        \n        vocab_size_chars = len(chars_from_text)\n        \n        print(\"Chars vocab size: \", vocab_size_chars, \"\\n\")\n        print(\"Sampled and cleaned chars from text: \", chars_from_text, \"\\n\\n\")\n        \n        # WORD PROCESSING\n        \n        text_lowered_split = text.lower().split()\n        words_from_text = sorted(list(set(text_lowered_split)))\n        \n        # Function to display data that might be misleading for our model.\n        unuseful_words = []\n        for word in words_from_text:\n            if len(word) < 4 and word not in unuseful_words:\n                unuseful_words.append(word)\n\n        print(\"Words that might be misleading for our model: \", unuseful_words, \" \\n\\n\")\n        \n        # Remove number containing words from the dataset and\n        # also strip words from punctuation and symbols.\n        words_from_text_tmp = []\n        for word in words_from_text:\n            word_to_add = ''\n            if not any(char.isdigit() for char in word):\n                word_to_add = word\n                #if (len(word_to_add) > 2 and word_to_add not in words_from_text_tmp):\n                words_from_text_tmp.append(word_to_add)\n                \n        words_from_text = words_from_text_tmp\n\n    vocab_size_words = len(words_from_text)\n    \n    print(\"Words vocab size: \", vocab_size_words, \"\\n\")\n    print(\"Sampled and cleaned words from text: \", words_from_text[:100])\n    \n    \n    # CHAR / WORD VOCAB SIZE DIFFERENCE VISUALIZATION\n\n    barplot_data = pd.DataFrame({\n        'Category': ['Words', 'Chars'],\n        'Count': [vocab_size_words, vocab_size_chars]\n    })\n    \n    plt.figure(figsize=(10,6))\n    plt.title(\"Vocab size difference for words and chars as tokens.\")\n    sns.barplot(x='Category', y='Count', data=barplot_data)\n    plt.ylabel(\"Number of tokens\")\n    plt.show()\n    \n    return vocab_size_words, vocab_size_chars, text, words_from_text, chars_from_text","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:02.230490Z","iopub.execute_input":"2024-09-30T13:05:02.231217Z","iopub.status.idle":"2024-09-30T13:05:02.244015Z","shell.execute_reply.started":"2024-09-30T13:05:02.231185Z","shell.execute_reply":"2024-09-30T13:05:02.243091Z"},"trusted":true},"execution_count":3,"outputs":[]},{"cell_type":"code","source":"vocab_size_words, vocab_size_chars, text, words_from_text, chars_from_text = load_clean_data()","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:05.651286Z","iopub.execute_input":"2024-09-30T13:05:05.651618Z","iopub.status.idle":"2024-09-30T13:05:05.927088Z","shell.execute_reply.started":"2024-09-30T13:05:05.651592Z","shell.execute_reply":"2024-09-30T13:05:05.926173Z"},"trusted":true},"execution_count":4,"outputs":[{"name":"stdout","text":"Chars vocab size:  30 \n\nSampled and cleaned chars from text:  [' ', '.', '!', '?', '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\n\nWords that might be misleading for our model:  ['\"a', '\"as', '\"do', '\"en', '\"i', '\"it', '\"m.', '\"my', '\"or', '\"to', '&c.', \"'a\", \"'as\", \"'be\", \"'do\", \"'he\", \"'i\", \"'if\", \"'in\", \"'it\", \"'m.\", \"'my\", \"'no\", \"'o\", \"'o,\", \"'of\", \"'to\", \"'we\", '(1)', '(2)', '(3)', '(4)', '(5)', '(6)', '(7)', '(8)', '(a', '(as', '(be', '(c.', '(he', '(if', '(in', '(is', '(of', '(or', '(to', '(v.', '1', '1,', '1.', '10.', '11', '12', '12.', '14', '140', '143', '145', '169', '17', '17,', '174', '19', '19.', '2', '2.', '21.', '22,', '22.', '250', '26,', '28', '28,', '287', '2nd', '3,', '3.', '31', '345', '350', '353', '37.', '373', '4.', '412', '44.', '46', '486', '4th', '5.', '512', '5th', '6,', '6.', '66', '6th', '7', '7,', '7.', '757', '79', '8.', '839', '898', '9th', '[1]', '[2]', '[3]', '[4]', '[5]', '[6]', '[7]', '[8]', '[9]', '_in', 'a', 'a.', 'a.,', 'act', 'ad', 'ad.', 'add', 'ado', 'age', 'ago', 'ah,', 'aid', 'aim', 'air', 'all', 'am', 'am,', 'an', 'and', 'any', 'apt', 'are', 'arm', 'art', 'as', 'as,', 'ask', 'at', 'at,', 'at.', 'at?', 'axe', 'b.', 'b.c', 'bad', 'bag', 'be', 'be)', 'be,', 'be.', 'be:', 'be;', 'be?', 'bed', 'bee', 'bet', 'bid', 'big', 'bit', 'boy', 'bud', 'but', 'by', 'by,', 'by;', 'c.', 'can', 'cap', 'cry', 'cut', 'd.', 'day', 'de', 'did', 'die', 'dig', 'dio', 'do', 'do!', 'do,', 'do.', 'do:', 'do;', 'do?', 'dog', 'don', 'dry', 'due', 'dye', 'ear', 'eat', 'ed.', 'end', 'ep.', 'et', 'eye', 'f.', 'far', 'few', 'fig', 'fit', 'fix', 'fl.', 'fly', 'foe', 'for', 'fro', 'g.', 'gar', 'get', 'go', 'go,', 'god', 'got', 'gr.', 'h.', 'had', 'hap', 'has', 'he', 'he\"', 'he)', 'he,', 'he.', 'he:', 'her', 'hid', 'him', 'his', 'hit', 'ho!', 'ho,', 'hot', 'how', 'hut', 'i', 'i)', 'i,', 'i.', 'i.,', 'i?', 'if', 'if,', 'ii', 'ii,', 'ii.', 'ill', 'imp', 'in', 'in,', 'in.', 'in;', 'in?', 'ir', 'is', 'is!', 'is)', 'is,', 'is.', 'is:', 'is;', 'it', 'it!', 'it)', 'it,', 'it.', 'it:', 'it;', 'it?', 'its', 'iv', 'iv.', 'ix.', 'j.', 'joy', 'kl', 'l.', 'law', 'lay', 'led', 'leg', 'let', 'li.', 'lie', 'lot', 'low', 'lv.', 'm', 'm,', 'm.', 'mad', 'mai', 'man', 'may', 'md', 'me', 'me!', 'me,', 'me.', 'me:', 'me;', 'me?', 'men', 'met', 'mh', 'mr.', 'ms.', 'my', 'n.', 'nay', 'net', 'new', 'no', 'no,', 'no.', 'no?', 'nor', 'not', 'now', 'o', 'odd', 'of', 'of,', 'of.', 'of:', 'of;', 'of?', 'off', 'oft', 'oh,', 'oil', 'old', 'on', 'on,', 'on.', 'on;', 'one', 'op.', 'or', 'or,', 'our', 'out', 'owe', 'own', 'ox,', 'p.', 'par', 'pay', 'pen', 'pig', 'put', 'q.', 'r.', 'ram', 'rid', 'rob', 'rod', 'run', 's.', 'sap', 'saw', 'say', 'sc.', 'see', 'set', 'she', 'sin', 'sit', 'six', 'sky', 'so', 'so)', 'so,', 'so.', 'so:', 'so;', 'so?', 'son', 'st.', 'sum', 'sun', 'ta', 'ten', 'the', 'thy', 'ti,', 'tin', 'tit', 'to', 'to,', 'too', 'top', 'try', 'two', 'up', 'up,', 'up.', 'us', 'us,', 'us.', 'us:', 'us;', 'us?', 'use', 'v', 'v.', 'vex', 'vi', 'vi.', 'vii', 'vow', 'war', 'was', 'wax', 'way', 'we', 'we:', 'wet', 'who', 'why', 'win', 'wit', 'woe', 'won', 'x', 'x.', 'xi', 'xi.', 'xii', 'xiv', 'xl.', 'xv.', 'xx.', 'ye', 'yea', 'yet', 'you', '§', 'μή,', 'μὲ.', 'τὸ', 'τῇ']  \n\n\nWords vocab size:  9959 \n\nSampled and cleaned words from text:  ['\"(and', '\"a', '\"affected', '\"agathos\"', '\"and', '\"as', '\"both', '\"but,', '\"cajeta\"', '\"chrysippus\"', '\"cithaeron\"', '\"claudius', '\"combined', '\"commonwealth\"', '\"congiaries\"', '\"consider,\"', '\"cut', '\"decree\"', '\"do', '\"do,', '\"doles.\"', '\"ears', '\"en', '\"epictetus\"', '\"for', '\"four', '\"frost\"', '\"gardens,\"', '\"harlot,\"', '\"high', '\"honour', '\"how', '\"i', '\"indifferent\"', '\"it', '\"layman,\"', '\"lives,\"', '\"m.', '\"man', '\"matter.\"', '\"my', '\"new', '\"one', '\"or', '\"pagus.\\'', '\"paltry', '\"patient', '\"phocion\"', '\"plato\"', '\"plato\":', '\"practical', '\"priest', '\"rhetoric\"', '\"rightness\"', '\"rigour.\"', '\"roarer\"', '\"says', '\"simple', '\"sixty\"', '\"straight,', '\"straightness.\"', '\"strain.\"', '\"strict', '\"suspension', '\"that', '\"the', '\"they', '\"thou', '\"to', '\"tragedian.\"', '\"us\"?', '\"whatever', '\"with', '\"wonder', '\"wood\":', '&c.', \"&c.'\", '&c.,', \"'a\", \"'affection'\", \"'after\", \"'all\", \"'and\", \"'as\", \"'as,\", \"'aurelius\", \"'be\", \"'but\", \"'cutting\", \"'do\", \"'doth\", \"'either\", \"'envy\", \"'for\", \"'fronto\", \"'he\", \"'holding\", \"'i\", \"'if\", \"'ill-tempered,'\"]\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"<Figure size 1000x600 with 1 Axes>","image/png":"iVBORw0KGgoAAAANSUhEUgAAA2QAAAIjCAYAAABswtioAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuNSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/xnp5ZAAAACXBIWXMAAA9hAAAPYQGoP6dpAABUvElEQVR4nO3de3zP9f//8ft7s5PZIWIHxzkbIhM5k8Nk5VhSK5IOMjmmSIkimUORUzqgQieHigzN5LTPSHxISEX00bbCNhtmttfvj757/bxt2Jv3emm7XS+X9+Wy9/P1fL3ej+drr723+16v1/NtMwzDEAAAAADgH+didQEAAAAAUFwRyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIANxUbDabBg8eXGjbX7RokWw2m44ePVpor1FQjz76qKpUqWLXZrPZNH78eLu2nTt3qnnz5vL29pbNZtOePXskSTExMWrYsKE8PT1ls9mUkpLyj9R9syrO++Po0aOy2WxatGiR1aU4JL/j/Xrkjn/atGk3XlQxV6VKFd1zzz1WlwEUKwQyoBjp2rWrSpYsqTNnzlyxT2RkpNzd3XXy5Ml/sDJcSVZWlu6//36dOnVKb7zxhj788ENVrlxZJ0+eVO/eveXl5aU5c+boww8/lLe3t9XlWob9AUjbt2/X+PHji9U/I4CioITVBQD450RGRuqrr77SypUr1bdv3zzLz549qy+++EKdO3dWmTJlLKiw8D3yyCPq06ePPDw8rC4lX+fOnVOJEv//rfmXX37Rb7/9pnfeeUePP/642R4TE6MzZ87o1VdfVYcOHawo9aayc+dO9geKve3bt2vChAl69NFH5e/vb3U5AAqIM2RAMdK1a1f5+Pho6dKl+S7/4osvlJGRocjIyH+4sn+Oq6ureUnbzcjT09MukCUnJ0tSnj+urtR+IzIyMpy2rX9acdgfZ8+etbqEYu9mOyYAFA0EMqAY8fLyUs+ePRUbG2v+AXuppUuXysfHR127dpUk/frrr7r//vtVunRplSxZUnfeeafWrFmTZ73z589r/Pjxqlmzpjw9PRUUFKSePXvql19+MftMmzZNzZs3V5kyZeTl5aWwsDB9/vnnV6x1yZIlqlWrljw9PRUWFqbNmzcXaIxvvfWW6tatq5IlS+qWW25R48aN7QLo5feQjR8/XjabLd/Ho48+aq6Xk5OjN998U3Xr1pWnp6cCAgL01FNP6fTp0wWqa9WqVapXr548PT1Vr149rVy5Mt9+l95T8+ijj6pNmzaSpPvvv182m01t27ZV27Zt1a9fP0nSHXfckafWhIQEde7cWX5+fipZsqTatGmjbdu22b1O7rh//PFHPfTQQ7rlllvUsmVLc/lHH32ksLAweXl5qXTp0urTp4+OHz9ut422bduqXr16+vHHH9WuXTuVLFlS5cuXV3R0dJ5xFeQYud59fK398dlnn5ljufXWW/Xwww/rf//7n902Hn30UZUqVUq//PKLunTpIh8fnyv+Y2Lv3r2y2Wz68ssvzbZdu3bJZrOpUaNGdn3vvvtuNW3a1K5t7ty5qlu3rjw8PBQcHKyoqKg8l5jl7ttdu3apdevWKlmypF544QVJUkpKih599FH5+fnJ399f/fr1y/cStcTERPXv318VKlSQh4eHgoKC1K1bt2veP7l37149+uijqlq1qjw9PRUYGKjHHnssz2XMucfQzz//bJ6R8fPzU//+/fOEx8zMTA0fPlxly5Y132N+//33q9ZxqYIcP7kWLFigatWqycPDQ3fccYd27tx5Q+PL72eksPftmTNnNGzYMFWpUkUeHh4qV66cOnbsqO+///6K2x4/frxGjRolSQoJCTHfx3Jrunjxol599VVz31SpUkUvvPCCMjMzr1qzJC1evFglSpQwty859j5TkGNkw4YNatmypfz9/VWqVCnVqlXLPOaBoo5LFoFiJjIyUosXL9ann35qN3nGqVOntG7dOj344IPy8vJSUlKSmjdvrrNnz2rIkCEqU6aMFi9erK5du+rzzz9Xjx49JEnZ2dm65557FBsbqz59+mjo0KE6c+aMNmzYoB9++EHVqlWTJM2cOVNdu3ZVZGSkLly4oI8//lj333+/Vq9erYiICLsav/32W33yyScaMmSIPDw8NHfuXHXu3Fk7duxQvXr1rji2d955R0OGDNF9992noUOH6vz589q7d68SEhL00EMP5btOz549Vb16dbu2Xbt26c0331S5cuXMtqeeekqLFi1S//79NWTIEB05ckSzZ8/W7t27tW3bNrm5uV2xrvXr16tXr14KDQ3V5MmTdfLkSfOPuat56qmnVL58eb322msaMmSI7rjjDgUEBEiSatWqpQULFuiVV15RSEiIuZ83btyou+++W2FhYXr55Zfl4uKihQsX6q677tKWLVvUpEkTu9e4//77VaNGDb322msyDEOSNGnSJL300kvq3bu3Hn/8cf35559666231Lp1a+3evdvuLNTp06fVuXNn9ezZU71799bnn3+u559/XvXr19fdd98tqeDHyPXu47Fjx15xf+Ru74477tDkyZOVlJSkmTNnatu2bXnGcvHiRYWHh6tly5aaNm2aSpYsme/r1atXT/7+/tq8ebP5z4stW7bIxcVF//3vf5WWliZfX1/l5ORo+/btevLJJ811x48frwkTJqhDhw56+umndejQIc2bN087d+7MM8aTJ0/q7rvvVp8+ffTwww8rICBAhmGoW7du2rp1qwYOHKg6depo5cqVZiC9VK9evbR//34988wzqlKlipKTk7VhwwYdO3Ysz2Qyl9qwYYN+/fVX9e/fX4GBgdq/f78WLFig/fv36z//+U+es8u9e/dWSEiIJk+erO+//17vvvuuypUrpylTpph9Hn/8cX300Ud66KGH1Lx5c23cuDHPz/2VFPT4kf7+p9KZM2f01FNPyWazKTo6Wj179tSvv/5q7ltHx5ffz0hh79uBAwfq888/1+DBgxUaGqqTJ09q69atOnDgQJ7Qn6tnz5766aeftGzZMr3xxhu69dZbJUlly5Y1vweLFy/Wfffdp5EjRyohIUGTJ0/WgQMHrvgPIunvgDtw4EC98MILmjhxoiTH32eudYzs379f99xzj2677Ta98sor8vDw0M8//5wn4AFFlgGgWLl48aIRFBRkNGvWzK59/vz5hiRj3bp1hmEYxrBhwwxJxpYtW8w+Z86cMUJCQowqVaoY2dnZhmEYxvvvv29IMmbMmJHntXJycsyvz549a7fswoULRr169Yy77rrLrl2SIcn47rvvzLbffvvN8PT0NHr06HHVsXXr1s2oW7fuVfssXLjQkGQcOXIk3+V//vmnUalSJaN+/fpGenq6YRiGsWXLFkOSsWTJEru+MTEx+bZfrmHDhkZQUJCRkpJitq1fv96QZFSuXNmuryTj5ZdfNp/HxcUZkozPPvss33Hs3LnTbMvJyTFq1KhhhIeH59n3ISEhRseOHc22l19+2ZBkPPjgg3bbPXr0qOHq6mpMmjTJrn3fvn1GiRIl7NrbtGljSDI++OADsy0zM9MIDAw0evXqZbYV5Bi50X2c3/64cOGCUa5cOaNevXrGuXPnzPbVq1cbkoxx48aZbf369TMkGaNHj77q6+SKiIgwmjRpYj7v2bOn0bNnT8PV1dVYu3atYRiG8f333xuSjC+++MIwDMNITk423N3djU6dOpk/P4ZhGLNnzzYkGe+//77Zlrtv58+fb/e6q1atMiQZ0dHRZtvFixeNVq1aGZKMhQsXGoZhGKdPnzYkGVOnTi3QeC51+c+qYRjGsmXLDEnG5s2bzbbcY+ixxx6z69ujRw+jTJky5vM9e/YYkoxBgwbZ9XvooYfyHO/5Kcjxc+TIEUOSUaZMGePUqVPm8i+++MKQZHz11VfXPb7Lf0b+iX3r5+dnREVFObz9qVOn5vv+lvs9ePzxx+3an332WUOSsXHjRrOtcuXKRkREhGEYhjFz5kzDZrMZr776qrn8et5nrnWMvPHGG4Yk488//3R4zEBRwCWLQDHj6uqqPn36KD4+3u7ymqVLlyogIEDt27eXJH399ddq0qSJ3WVspUqV0pNPPqmjR4/qxx9/lCQtX75ct956q5555pk8r3Xpf5q9vLzMr0+fPq3U1FS1atUq30twmjVrprCwMPN5pUqV1K1bN61bt07Z2dlXHJu/v79+//33PJcoFVR2drYefPBBnTlzRitXrjRn6fvss8/k5+enjh076q+//jIfYWFhKlWqlOLi4q64zT/++EN79uxRv3795OfnZ7Z37NhRoaGh11XnlezZs0eHDx/WQw89pJMnT5p1ZmRkqH379tq8ebNycnLs1hk4cKDd8xUrVignJ0e9e/e2G2tgYKBq1KiRZ6ylSpXSww8/bD53d3dXkyZN9Ouvv5ptBTlGbmQfX8l3332n5ORkDRo0SJ6enmZ7RESEateune/lt08//XSBtp177ObeU7R161Z16dJFDRs21JYtWyT9fdbMZrOZP0PffPONLly4oGHDhsnF5f//+n3iiSfk6+ubpx4PDw/179/fru3rr79WiRIl7Op0dXXNs2+9vLzk7u6uTZs2Ffiy2kvXzXX+/Hn99ddfuvPOOyUp35/Xy4+hVq1a6eTJk0pLSzNrlqQhQ4bY9Rs2bFiB6inoe4wkPfDAA7rlllvsapFkdzze6Pj+iX3r7++vhIQEnThxwqHtX0nu92DEiBF27SNHjpSkfH8WoqOjNXToUE2ZMkUvvvii2e6M95nLj5HcM9VffPFFnnWB4oBABhRDuffG5N5b9fvvv2vLli3q06ePXF1dJUm//fabatWqlWfdOnXqmMulv2cBrFWrlt1EFPlZvXq17rzzTnl6eqp06dIqW7as5s2bp9TU1Dx9a9SokaetZs2aOnv2rP78888rvsbzzz+vUqVKqUmTJqpRo4aioqIcuuTlxRdf1MaNG7V06VK7y6AOHz6s1NRUlStXTmXLlrV7pKen53s/Xq7c/ZTfmPLbvzfi8OHDkqR+/frlqfPdd99VZmZmnv0dEhKSZxuGYahGjRp5tnHgwIE8Y61QoUKeP4pvueUWuz9UC3KM3Mg+vpLcfZ/ffq5du7a5PFeJEiWueRlprlatWunixYuKj4/XoUOHlJycrFatWql169Z2gSw0NFSlS5e+aj3u7u6qWrVqnnrKly8vd3f3PGMKCgpSqVKl7Nov36aHh4emTJmitWvXKiAgQK1bt1Z0dLQSExOvObZTp05p6NChCggIkJeXl8qWLWseJ/n9vFaqVMnueW4gyj0GfvvtN7m4uNj9TOVX85UU9D2mILVIjo/v8p+Rf2LfRkdH64cfflDFihXVpEkTjR8/3i5UOir3e3D55dmBgYHy9/fPc+x9++23ev755/X888/b3TcmXd/7zLW+Lw888IBatGihxx9/XAEBAerTp48+/fRTwhmKDe4hA4qhsLAw1a5dW8uWLdMLL7ygZcuWyTCMQptdccuWLeratatat26tuXPnKigoSG5ublq4cOEVZ3y8HnXq1NGhQ4e0evVqxcTEaPny5Zo7d67GjRunCRMmXHXdVatWacqUKXr11VfVuXNnu2U5OTkqV66clixZku+6ufdoWC33j5epU6eqYcOG+fa5/A/5S/9jn7sNm82mtWvXmuH8auvn10eSea9NQd0M+9jDw8PuzNXVNG7cWJ6entq8ebMqVaqkcuXKqWbNmmrVqpXmzp2rzMxMbdmyxbzX8npc/r1x1LBhw3Tvvfdq1apVWrdunV566SVNnjxZGzdu1O23337F9Xr37q3t27dr1KhRatiwoUqVKqWcnBx17tw53z+QnXUMOENBanF0fPl9Hwp73/bu3VutWrXSypUrtX79ek2dOlVTpkzRihUrzHszr0dBZ5etW7euUlJS9OGHH+qpp56yC6XX8z5zre+Ll5eXNm/erLi4OK1Zs0YxMTH65JNPdNddd2n9+vVXXB8oKghkQDEVGRmpl156SXv37tXSpUtVo0YN3XHHHebyypUr69ChQ3nWO3jwoLlckqpVq6aEhARlZWVdcdKF5cuXy9PTU+vWrbP7/K+FCxfm2z/3P7CX+umnn1SyZMlr/mHu7e2tBx54QA888IAuXLignj17atKkSRozZozdZWuXb7tfv37q3r17vrN6VatWTd98841atGjh8B/JufspvzHlt39vRO4ZCF9f3+v+LK5q1arJMAyFhISoZs2aTqvrWsfIjezjK8nd94cOHdJdd91lt+zQoUPm8uuRe2nmli1bVKlSJfPSuFatWikzM1NLlixRUlKSWrdunW89VatWNdsvXLigI0eOFOh7VrlyZcXGxio9Pd3uj94rHUvVqlXTyJEjNXLkSB0+fFgNGzbU9OnT9dFHH+Xb//Tp04qNjdWECRM0btw4sz2/47egKleurJycHPNM17Vqzm8M1zp+CsqZ4yvsfRsUFKRBgwZp0KBBSk5OVqNGjTRp0qSrBrIrBa7c78Hhw4fNqxwkKSkpSSkpKXl+Fm699VZ9/vnnatmypdq3b6+tW7cqODjYHLd0Y+8z+XFxcVH79u3Vvn17zZgxQ6+99prGjh2ruLg4PlsQRR6XLALFVO7ZsHHjxmnPnj15zo516dJFO3bsUHx8vNmWkZGhBQsWqEqVKub9T7169dJff/2l2bNn53mN3P9+urq6ymaz2d3/dfToUa1atSrf2uLj4+3upzh+/Li++OILderU6ar/Kb186mh3d3eFhobKMAxlZWXlu056erp69Oih8uXLa/Hixfn+QdO7d29lZ2fr1VdfzbPs4sWL+U45nisoKEgNGzbU4sWL7S7j2bBhg3kfnrOEhYWpWrVqmjZtmtLT0/Msv9rlnrl69uwpV1dXTZgwIc8ZDsMw8uzjgijIMXIj+/hKGjdurHLlymn+/Pl2U3uvXbtWBw4cKPAsf1fSqlUrJSQkKC4uzgxkt956q+rUqWPOHpfbLkkdOnSQu7u7Zs2aZbdv33vvPaWmphaoni5duujixYuaN2+e2Zadna233nrLrt/Zs2d1/vx5u7Zq1arJx8fnqtOc5/58Xf69f/PNN69Z25XkBohZs2Zd1zYLcvwUlDPGV9j7Njs7O88lf+XKlVNwcPA1p6jPve/18p+XLl265PtaM2bMkKR8j70KFSrom2++0blz59SxY0fzZ98Z7zOXO3XqVJ623LNvl4754MGDOnbsmMPbB252nCEDiqmQkBA1b95cX3zxhSTlCWSjR4/WsmXLdPfdd2vIkCEqXbq0Fi9erCNHjmj58uXmpV19+/bVBx98oBEjRmjHjh1q1aqVMjIy9M0332jQoEHq1q2bIiIiNGPGDHXu3FkPPfSQkpOTNWfOHFWvXl179+7NU1u9evUUHh5uN+29pGtedtipUycFBgaqRYsWCggI0IEDBzR79mxFRETIx8cn33UmTJigH3/8US+++KK5L3JVq1ZNzZo1U5s2bfTUU09p8uTJ2rNnjzp16iQ3NzcdPnxYn332mWbOnKn77rvvinVNnjxZERERatmypR577DGdOnXK/Ly0/P6guV4uLi569913dffdd6tu3brq37+/ypcvr//973+Ki4uTr6+vvvrqq6tuo1q1apo4caLGjBmjo0ePqnv37vLx8dGRI0e0cuVKPfnkk3r22Wcdqqsgx8iN7uP8uLm5acqUKerfv7/atGmjBx980Jz2vkqVKho+fLhD27tcq1atNGnSJB0/ftwueLVu3Vpvv/22qlSpYndPWtmyZTVmzBhNmDBBnTt3VteuXXXo0CHNnTtXd9xxh93kKFdy7733qkWLFho9erSOHj2q0NBQrVixIs8f8D/99JPat2+v3r17KzQ0VCVKlNDKlSuVlJSkPn36XHH7vr6+5j1RWVlZKl++vNavX68jR45cxx76W8OGDfXggw9q7ty5Sk1NVfPmzRUbG6uff/65QOsX5PgpKGeMr7D37ZkzZ1ShQgXdd999atCggUqVKqVvvvlGO3fu1PTp069aW+5kSGPHjlWfPn3k5uame++9Vw0aNFC/fv20YMECpaSkqE2bNtqxY4cWL16s7t27q127dvlur3r16lq/fr3atm2r8PBwbdy4Ub6+vjf8PnO5V155RZs3b1ZERIQqV66s5ORkzZ07VxUqVLCbWKpOnTpq06aNNm3a5ND2gZvePz6vI4Cbxpw5cwxJdtN3X+qXX34x7rvvPsPf39/w9PQ0mjRpYqxevTpPv7Nnzxpjx441QkJCDDc3NyMwMNC47777jF9++cXs89577xk1atQwPDw8jNq1axsLFy40p0S+lCQjKirK+Oijj8z+t99+uxEXF3fN8bz99ttG69atjTJlyhgeHh5GtWrVjFGjRhmpqalmn8unvc+d7jy/R79+/ey2v2DBAiMsLMzw8vIyfHx8jPr16xvPPfecceLEiWvWtnz5cqNOnTqGh4eHERoaaqxYscLo16+fU6e9z7V7926jZ8+e5n6oXLmy0bt3byM2Ntbsk7vvrzTN9PLly42WLVsa3t7ehre3t1G7dm0jKirKOHTokNmnTZs2+X7MQH7jKsgxYhjXv4+vtj8++eQT4/bbbzc8PDyM0qVLG5GRkcbvv/+ep2Zvb++rvsbl0tLSDFdXV8PHx8e4ePGi2f7RRx8ZkoxHHnkk3/Vmz55t1K5d23BzczMCAgKMp59+2jh9+rRdnyvtW8MwjJMnTxqPPPKI4evra/j5+RmPPPKIsXv3brtp7//66y8jKirKqF27tuHt7W34+fkZTZs2NT799NNrjuv33383evToYfj7+xt+fn7G/fffb5w4cSLPsXmlYyi/j5Y4d+6cMWTIEKNMmTKGt7e3ce+99xrHjx8v0LT3hnHt4yd32vv8pqK//DVudHyFvW8zMzONUaNGGQ0aNDB8fHwMb29vo0GDBsbcuXOvuX3DMIxXX33VKF++vOHi4mL3fcjKyjImTJhg7sOKFSsaY8aMMc6fP2+3/qXT3udKSEgwfHx8jNatW5tT99/I+8zlx0hsbKzRrVs3Izg42HB3dzeCg4ONBx980Pjpp5/s1pNktGnTpkD7Afg3sRmGBXfdAgAAAAC4hwwAAAAArEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAi/DB0E6Sk5OjEydOyMfHRzabzepyAAAAAFjEMAydOXNGwcHBcnG5+jkwApmTnDhxQhUrVrS6DAAAAAA3iePHj6tChQpX7UMgcxIfHx9Jf+90X19fi6sBAAAAYJW0tDRVrFjRzAhXQyBzktzLFH19fQlkAAAAAAp0KxOTegAAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARSwNZJs3b9a9996r4OBg2Ww2rVq1ym65YRgaN26cgoKC5OXlpQ4dOujw4cN2fU6dOqXIyEj5+vrK399fAwYMUHp6ul2fvXv3qlWrVvL09FTFihUVHR2dp5bPPvtMtWvXlqenp+rXr6+vv/7a6eMFAAAAgEtZGsgyMjLUoEEDzZkzJ9/l0dHRmjVrlubPn6+EhAR5e3srPDxc58+fN/tERkZq//792rBhg1avXq3NmzfrySefNJenpaWpU6dOqly5snbt2qWpU6dq/PjxWrBggdln+/btevDBBzVgwADt3r1b3bt3V/fu3fXDDz8U3uABAAAAFHs2wzAMq4uQJJvNppUrV6p79+6S/j47FhwcrJEjR+rZZ5+VJKWmpiogIECLFi1Snz59dODAAYWGhmrnzp1q3LixJCkmJkZdunTR77//ruDgYM2bN09jx45VYmKi3N3dJUmjR4/WqlWrdPDgQUnSAw88oIyMDK1evdqs584771TDhg01f/78AtWflpYmPz8/paamytfX11m7BQAAAMC/jCPZ4Ka9h+zIkSNKTExUhw4dzDY/Pz81bdpU8fHxkqT4+Hj5+/ubYUySOnToIBcXFyUkJJh9WrdubYYxSQoPD9ehQ4d0+vRps8+lr5PbJ/d18pOZmam0tDS7BwAAAAA44qYNZImJiZKkgIAAu/aAgABzWWJiosqVK2e3vESJEipdurRdn/y2celrXKlP7vL8TJ48WX5+fuajYsWKjg4RAAAAQDF30waym92YMWOUmppqPo4fP251SQAAAAD+ZUpYXcCVBAYGSpKSkpIUFBRkticlJalhw4Zmn+TkZLv1Ll68qFOnTpnrBwYGKikpya5P7vNr9cldnh8PDw95eHhcx8isETbqA6tLAACn2jW1r9UlAABww27aM2QhISEKDAxUbGys2ZaWlqaEhAQ1a9ZMktSsWTOlpKRo165dZp+NGzcqJydHTZs2Nfts3rxZWVlZZp8NGzaoVq1auuWWW8w+l75Obp/c1wEAAACAwmBpIEtPT9eePXu0Z88eSX9P5LFnzx4dO3ZMNptNw4YN08SJE/Xll19q37596tu3r4KDg82ZGOvUqaPOnTvriSee0I4dO7Rt2zYNHjxYffr0UXBwsCTpoYcekru7uwYMGKD9+/frk08+0cyZMzVixAizjqFDhyomJkbTp0/XwYMHNX78eH333XcaPHjwP71LAAAAABQjll6y+N1336ldu3bm89yQ1K9fPy1atEjPPfecMjIy9OSTTyolJUUtW7ZUTEyMPD09zXWWLFmiwYMHq3379nJxcVGvXr00a9Ysc7mfn5/Wr1+vqKgohYWF6dZbb9W4cePsPqusefPmWrp0qV588UW98MILqlGjhlatWqV69er9A3sBAAAAQHF103wO2b/dzf45ZNxDBqCo4R4yAMDNqkh8DhkAAAAAFHUEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAsQiADAAAAAIsQyAAAAADAIgQyAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiN3Ugy87O1ksvvaSQkBB5eXmpWrVqevXVV2UYhtnHMAyNGzdOQUFB8vLyUocOHXT48GG77Zw6dUqRkZHy9fWVv7+/BgwYoPT0dLs+e/fuVatWreTp6amKFSsqOjr6HxkjAAAAgOLrpg5kU6ZM0bx58zR79mwdOHBAU6ZMUXR0tN566y2zT3R0tGbNmqX58+crISFB3t7eCg8P1/nz580+kZGR2r9/vzZs2KDVq1dr8+bNevLJJ83laWlp6tSpkypXrqxdu3Zp6tSpGj9+vBYsWPCPjhcAAABA8VLC6gKuZvv27erWrZsiIiIkSVWqVNGyZcu0Y8cOSX+fHXvzzTf14osvqlu3bpKkDz74QAEBAVq1apX69OmjAwcOKCYmRjt37lTjxo0lSW+99Za6dOmiadOmKTg4WEuWLNGFCxf0/vvvy93dXXXr1tWePXs0Y8YMu+AGAAAAAM50U58ha968uWJjY/XTTz9Jkv773/9q69atuvvuuyVJR44cUWJiojp06GCu4+fnp6ZNmyo+Pl6SFB8fL39/fzOMSVKHDh3k4uKihIQEs0/r1q3l7u5u9gkPD9ehQ4d0+vTpfGvLzMxUWlqa3QMAAAAAHHFTnyEbPXq00tLSVLt2bbm6uio7O1uTJk1SZGSkJCkxMVGSFBAQYLdeQECAuSwxMVHlypWzW16iRAmVLl3ark9ISEiebeQuu+WWW/LUNnnyZE2YMMEJowQAAABQXN3UZ8g+/fRTLVmyREuXLtX333+vxYsXa9q0aVq8eLHVpWnMmDFKTU01H8ePH7e6JAAAAAD/Mjf1GbJRo0Zp9OjR6tOnjySpfv36+u233zR58mT169dPgYGBkqSkpCQFBQWZ6yUlJalhw4aSpMDAQCUnJ9tt9+LFizp16pS5fmBgoJKSkuz65D7P7XM5Dw8PeXh43PggAQAAABRbN/UZsrNnz8rFxb5EV1dX5eTkSJJCQkIUGBio2NhYc3laWpoSEhLUrFkzSVKzZs2UkpKiXbt2mX02btyonJwcNW3a1OyzefNmZWVlmX02bNigWrVq5Xu5IgAAAAA4w00dyO69915NmjRJa9as0dGjR7Vy5UrNmDFDPXr0kCTZbDYNGzZMEydO1Jdffql9+/apb9++Cg4OVvfu3SVJderUUefOnfXEE09ox44d2rZtmwYPHqw+ffooODhYkvTQQw/J3d1dAwYM0P79+/XJJ59o5syZGjFihFVDBwAAAFAM3NSXLL711lt66aWXNGjQICUnJys4OFhPPfWUxo0bZ/Z57rnnlJGRoSeffFIpKSlq2bKlYmJi5OnpafZZsmSJBg8erPbt28vFxUW9evXSrFmzzOV+fn5av369oqKiFBYWpltvvVXjxo1jynsAAAAAhcpmGIZhdRFFQVpamvz8/JSamipfX1+ry8kjbNQHVpcAAE61a2pfq0sAACBfjmSDm/qSRQAAAAAoyghkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFjkhgNZWlqaVq1apQMHDjijHgAAAAAoNhwOZL1799bs2bMlSefOnVPjxo3Vu3dv3XbbbVq+fLnTCwQAAACAosrhQLZ582a1atVKkrRy5UoZhqGUlBTNmjVLEydOdHqBAAAAAFBUORzIUlNTVbp0aUlSTEyMevXqpZIlSyoiIkKHDx92eoEAAAAAUFQ5HMgqVqyo+Ph4ZWRkKCYmRp06dZIknT59Wp6enk4vEAAAAACKqhKOrjBs2DBFRkaqVKlSqly5stq2bSvp70sZ69ev7+z6AAAAAKDIcjiQDRo0SE2aNNHx48fVsWNHubj8fZKtatWq3EMGAAAAAA5wOJBJUuPGjdW4cWO7toiICKcUBAAAAADFhcOBLDs7W4sWLVJsbKySk5OVk5Njt3zjxo1OKw4AAAAAijKHA9nQoUO1aNEiRUREqF69erLZbIVRFwAAAAAUeQ4Hso8//liffvqpunTpUhj1AAAAAECx4fC09+7u7qpevXph1AIAAAAAxYrDgWzkyJGaOXOmDMMojHoAAAAAoNhw+JLFrVu3Ki4uTmvXrlXdunXl5uZmt3zFihVOKw4AAAAAijKHA5m/v7969OhRGLUAAAAAQLHicCBbuHBhYdQBAAAAAMWOw/eQSdLFixf1zTff6O2339aZM2ckSSdOnFB6erpTiwMAAACAoszhM2S//fabOnfurGPHjikzM1MdO3aUj4+PpkyZoszMTM2fP78w6gQAAACAIsfhM2RDhw5V48aNdfr0aXl5eZntPXr0UGxsrFOLAwAAAICizOEzZFu2bNH27dvl7u5u116lShX973//c1phAAAAAFDUOXyGLCcnR9nZ2Xnaf//9d/n4+DilKAAAAAAoDhwOZJ06ddKbb75pPrfZbEpPT9fLL7+sLl26OLM2AAAAACjSHL5kcfr06QoPD1doaKjOnz+vhx56SIcPH9att96qZcuWFUaNAAAAAFAkOXyGrEKFCvrvf/+rsWPHavjw4br99tv1+uuva/fu3SpXrpzTC/zf//6nhx9+WGXKlJGXl5fq16+v7777zlxuGIbGjRunoKAgeXl5qUOHDjp8+LDdNk6dOqXIyEj5+vrK399fAwYMyDNF/969e9WqVSt5enqqYsWKio6OdvpYAAAAAOBSDgeyZcuWqUSJEoqMjFR0dLTmzp2rxx9/XF5eXho1apRTizt9+rRatGghNzc3rV27Vj/++KOmT5+uW265xewTHR2tWbNmaf78+UpISJC3t7fCw8N1/vx5s09kZKT279+vDRs2aPXq1dq8ebOefPJJc3laWpo6deqkypUra9euXZo6darGjx+vBQsWOHU8AAAAAHApm2EYhiMr+Pv7a9myZbr77rvt2ocPH66PP/5Yf/zxh9OKGz16tLZt26YtW7bku9wwDAUHB2vkyJF69tlnJUmpqakKCAjQokWL1KdPHx04cEChoaHauXOnGjduLEmKiYlRly5d9Pvvvys4OFjz5s3T2LFjlZiYaM4eOXr0aK1atUoHDx4sUK1paWny8/NTamqqfH19nTB65wob9YHVJQCAU+2a2tfqEgAAyJcj2cDhM2RLlizRgw8+qK1bt5ptzzzzjD799FPFxcU5Xu1VfPnll2rcuLHuv/9+lStXTrfffrveeecdc/mRI0eUmJioDh06mG1+fn5q2rSp4uPjJUnx8fHy9/c3w5gkdejQQS4uLkpISDD7tG7d2m4q//DwcB06dEinT5/Ot7bMzEylpaXZPQAAAADAEQ4HsoiICM2dO1ddu3bVrl27NGjQIK1YsUJxcXGqXbu2U4v79ddfNW/ePNWoUUPr1q3T008/rSFDhmjx4sWSpMTERElSQECA3XoBAQHmssTExDz3tpUoUUKlS5e265PfNi59jctNnjxZfn5+5qNixYo3OFoAAAAAxY3DsyxK0kMPPaSUlBS1aNFCZcuW1bfffqvq1as7uzbl5OSocePGeu211yRJt99+u3744QfNnz9f/fr1c/rrOWLMmDEaMWKE+TwtLY1QBgAAAMAhBQpklwaPS5UtW1aNGjXS3LlzzbYZM2Y4pzJJQUFBCg0NtWurU6eOli9fLkkKDAyUJCUlJSkoKMjsk5SUpIYNG5p9kpOT7bZx8eJFnTp1ylw/MDBQSUlJdn1yn+f2uZyHh4c8PDyuc2QAAAAAUMBAtnv37nzbq1evrrS0NHO5zWZzXmWSWrRooUOHDtm1/fTTT6pcubIkKSQkRIGBgYqNjTUDWFpamhISEvT0009Lkpo1a6aUlBTt2rVLYWFhkqSNGzcqJydHTZs2NfuMHTtWWVlZcnNzkyRt2LBBtWrVspvREQAAAACcqUCBzNmTdRTU8OHD1bx5c7322mvq3bu3duzYoQULFpjT0dtsNg0bNkwTJ05UjRo1FBISopdeeknBwcHq3r27pL/PqHXu3FlPPPGE5s+fr6ysLA0ePFh9+vRRcHCwpL8vwZwwYYIGDBig559/Xj/88INmzpypN954w5JxAwAAACgerusesly///67pL8/LLow3HHHHVq5cqXGjBmjV155RSEhIXrzzTcVGRlp9nnuueeUkZGhJ598UikpKWrZsqViYmLk6elp9lmyZIkGDx6s9u3by8XFRb169dKsWbPM5X5+flq/fr2ioqIUFhamW2+9VePGjbP7rDIAAAAAcDaHP4csJydHEydO1PTp05Weni5J8vHx0ciRIzV27Fi5uDg8cWORwOeQAcA/i88hAwDcrBzJBg6fIRs7dqzee+89vf7662rRooUkaevWrRo/frzOnz+vSZMmXV/VAAAAAFDMOBzIFi9erHfffVddu3Y122677TaVL19egwYNIpABAAAAQAE5fH3hqVOn8v0A6Nq1a+vUqVNOKQoAAAAAigOHA1mDBg00e/bsPO2zZ89WgwYNnFIUAAAAABQHDl+yGB0drYiICH3zzTdq1qyZJCk+Pl7Hjx/X119/7fQCAQAAAKCocvgMWZs2bfTTTz+pR48eSklJUUpKinr27KlDhw6pVatWhVEjAAAAABRJDp8hO3bsmCpWrJjv5B3Hjh1TpUqVnFIYAAAAABR1Dp8hCwkJ0Z9//pmn/eTJkwoJCXFKUQAAAABQHDgcyAzDkM1my9Oenp4uT09PpxQFAAAAAMVBgS9ZHDFihCTJZrPppZdeUsmSJc1l2dnZSkhIUMOGDZ1eIAAAAAAUVQUOZLt375b09xmyffv2yd3d3Vzm7u6uBg0a6Nlnn3V+hQAAAABQRBU4kMXFxUmS+vfvr5kzZ8rX17fQigIAAACA4sDhWRYXLlxYGHUAAAAAQLHj8KQeAAAAAADnIJABAAAAgEUIZAAAAABgkQIFskaNGun06dOSpFdeeUVnz54t1KIAAAAAoDgoUCA7cOCAMjIyJEkTJkxQenp6oRYFAAAAAMVBgWZZbNiwofr376+WLVvKMAxNmzZNpUqVyrfvuHHjnFogAAAAABRVBQpkixYt0ssvv6zVq1fLZrNp7dq1KlEi76o2m41ABgAAAAAFVKBAVqtWLX388ceSJBcXF8XGxqpcuXKFWhgAAAAAFHUOfzB0Tk5OYdQBAAAAAMWOw4FMkn755Re9+eabOnDggCQpNDRUQ4cOVbVq1ZxaHAAAAAAUZQ5/Dtm6desUGhqqHTt26LbbbtNtt92mhIQE1a1bVxs2bCiMGgEAAACgSHL4DNno0aM1fPhwvf7663nan3/+eXXs2NFpxQEAAABAUebwGbIDBw5owIABedofe+wx/fjjj04pCgAAAACKA4cDWdmyZbVnz5487Xv27GHmRQAAAABwgMOXLD7xxBN68skn9euvv6p58+aSpG3btmnKlCkaMWKE0wsEAAAAgKLK4UD20ksvycfHR9OnT9eYMWMkScHBwRo/fryGDBni9AIBAAAAoKhyOJDZbDYNHz5cw4cP15kzZyRJPj4+Ti8MAAAAAIq66/ocslwEMQAAAAC4fg5P6gEAAAAAcA4CGQAAAABYhEAGAAAAABZxKJBlZWWpffv2Onz4cGHVAwAAAADFhkOBzM3NTXv37i2sWgAAAACgWHH4ksWHH35Y7733XmHUAgAAAADFisPT3l+8eFHvv/++vvnmG4WFhcnb29tu+YwZM5xWHAAAAAAUZQ4Hsh9++EGNGjWSJP300092y2w2m3OqAgAAAIBiwOFAFhcXVxh1AAAAAECxc93T3v/8889at26dzp07J0kyDMNpRQEAAABAceBwIDt58qTat2+vmjVrqkuXLvrjjz8kSQMGDNDIkSOdXiAAAAAAFFUOB7Lhw4fLzc1Nx44dU8mSJc32Bx54QDExMU4tDgAAAACKMofvIVu/fr3WrVunChUq2LXXqFFDv/32m9MKAwAAAICizuEzZBkZGXZnxnKdOnVKHh4eTikKAAAAAIoDhwNZq1at9MEHH5jPbTabcnJyFB0drXbt2jm1OAAAAAAoyhy+ZDE6Olrt27fXd999pwsXLui5557T/v37derUKW3btq0wagQAAACAIsnhM2T16tXTTz/9pJYtW6pbt27KyMhQz549tXv3blWrVq0wagQAAACAIsnhM2SS5Ofnp7Fjxzq7FgAAAAAoVq4rkJ0+fVrvvfeeDhw4IEkKDQ1V//79Vbp0aacWBwAAAABFmcOXLG7evFlVqlTRrFmzdPr0aZ0+fVqzZs1SSEiINm/eXBg1AgAAAECR5PAZsqioKD3wwAOaN2+eXF1dJUnZ2dkaNGiQoqKitG/fPqcXCQAAAABFkcNnyH7++WeNHDnSDGOS5OrqqhEjRujnn392anEAAAAAUJQ5HMgaNWpk3jt2qQMHDqhBgwZOKQoAAAAAioMCXbK4d+9e8+shQ4Zo6NCh+vnnn3XnnXdKkv7zn/9ozpw5ev311wunSgAAAAAogmyGYRjX6uTi4iKbzaZrdbXZbMrOznZacf8maWlp8vPzU2pqqnx9fa0uJ4+wUR9YXQIAONWuqX2tLgEAgHw5kg0KdIbsyJEjTikMAAAAAPD/FSiQVa5cubDrAAAAAIBi57o+GPrEiRPaunWrkpOTlZOTY7dsyJAhTikMAAAAAIo6hwPZokWL9NRTT8nd3V1lypSRzWYzl9lsNgIZAAAAABSQw4HspZde0rhx4zRmzBi5uDg8az4AAAAA4P84nKjOnj2rPn36EMYAAAAA4AY5nKoGDBigzz77rDBqAQAAAIBixeFLFidPnqx77rlHMTExql+/vtzc3OyWz5gxw2nFAQAAAEBRdl2BbN26dapVq5Yk5ZnUAwAAAABQMA4HsunTp+v999/Xo48+WgjlAAAAAEDx4fA9ZB4eHmrRokVh1AIAAAAAxYrDgWzo0KF66623CqMWAAAAAChWHL5kcceOHdq4caNWr16tunXr5pnUY8WKFU4rDgAAAACKMocDmb+/v3r27FkYtQAAAABAseJwIFu4cGFh1AEAAAAAxY7D95ABAAAAAJzD4TNkISEhV/28sV9//fWGCgIAAACA4sLhQDZs2DC751lZWdq9e7diYmI0atQoZ9UFAAAAAEWew4Fs6NCh+bbPmTNH33333Q0XBAAAAADFhdPuIbv77ru1fPlyZ20OAAAAAIo8pwWyzz//XKVLl3bW5gAAAACgyHM4kN1+++1q1KiR+bj99tsVFBSkF154QS+88EJh1Gh6/fXXZbPZ7O5jO3/+vKKiolSmTBmVKlVKvXr1UlJSkt16x44dU0REhEqWLKly5cpp1KhRunjxol2fTZs2qVGjRvLw8FD16tW1aNGiQh0LAAAAADh8D1n37t3tnru4uKhs2bJq27atateu7ay68ti5c6fefvtt3XbbbXbtw4cP15o1a/TZZ5/Jz89PgwcPVs+ePbVt2zZJUnZ2tiIiIhQYGKjt27frjz/+UN++feXm5qbXXntNknTkyBFFRERo4MCBWrJkiWJjY/X4448rKChI4eHhhTYmAAAAAMWbzTAMw+oiriU9PV2NGjXS3LlzNXHiRDVs2FBvvvmmUlNTVbZsWS1dulT33XefJOngwYOqU6eO4uPjdeedd2rt2rW65557dOLECQUEBEiS5s+fr+eff15//vmn3N3d9fzzz2vNmjX64YcfzNfs06ePUlJSFBMTU6Aa09LS5Ofnp9TUVPn6+jp/J9ygsFEfWF0CADjVrql9rS4BAIB8OZIN/hUfDB0VFaWIiAh16NDBrn3Xrl3Kysqya69du7YqVaqk+Ph4SVJ8fLzq169vhjFJCg8PV1pamvbv32/2uXzb4eHh5jbyk5mZqbS0NLsHAAAAADiiwJcsuri4XPUDoSXJZrPluTfrRn388cf6/vvvtXPnzjzLEhMT5e7uLn9/f7v2gIAAJSYmmn0uDWO5y3OXXa1PWlqazp07Jy8vrzyvPXnyZE2YMOG6xwUAAAAABQ5kK1euvOKy+Ph4zZo1Szk5OU4pKtfx48c1dOhQbdiwQZ6enk7d9o0aM2aMRowYYT5PS0tTxYoVLawIAAAAwL9NgQNZt27d8rQdOnRIo0eP1ldffaXIyEi98sorTi1u165dSk5OVqNGjcy27Oxsbd68WbNnz9a6det04cIFpaSk2J0lS0pKUmBgoCQpMDBQO3bssNtu7iyMl/a5fGbGpKQk+fr65nt2TJI8PDzk4eFxw2MEAAAAUHxd1z1kJ06c0BNPPKH69evr4sWL2rNnjxYvXqzKlSs7tbj27dtr37592rNnj/lo3LixIiMjza/d3NwUGxtrrnPo0CEdO3ZMzZo1kyQ1a9ZM+/btU3Jystlnw4YN8vX1VWhoqNnn0m3k9sndBgAAAAAUBoemvU9NTdVrr72mt956Sw0bNlRsbKxatWpVWLXJx8dH9erVs2vz9vZWmTJlzPYBAwZoxIgRKl26tHx9ffXMM8+oWbNmuvPOOyVJnTp1UmhoqB555BFFR0crMTFRL774oqKioswzXAMHDtTs2bP13HPP6bHHHtPGjRv16aefas2aNYU2NgAAAAAocCCLjo7WlClTFBgYqGXLluV7CaMV3njjDbm4uKhXr17KzMxUeHi45s6day53dXXV6tWr9fTTT6tZs2by9vZWv3797C6vDAkJ0Zo1azR8+HDNnDlTFSpU0LvvvstnkAEAAAAoVAX+HDIXFxd5eXmpQ4cOcnV1vWK/FStWOK24fxM+hwwA/ll8DhkA4GblSDYo8Bmyvn37XnPaewAAAABAwRU4kC1atKgQywAAAACA4ue6ZlkEAAAAANw4AhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWuakD2eTJk3XHHXfIx8dH5cqVU/fu3XXo0CG7PufPn1dUVJTKlCmjUqVKqVevXkpKSrLrc+zYMUVERKhkyZIqV66cRo0apYsXL9r12bRpkxo1aiQPDw9Vr15dixYtKuzhAQAAACjmbupA9u233yoqKkr/+c9/tGHDBmVlZalTp07KyMgw+wwfPlxfffWVPvvsM3377bc6ceKEevbsaS7Pzs5WRESELly4oO3bt2vx4sVatGiRxo0bZ/Y5cuSIIiIi1K5dO+3Zs0fDhg3T448/rnXr1v2j4wUAAABQvNgMwzCsLqKg/vzzT5UrV07ffvutWrdurdTUVJUtW1ZLly7VfffdJ0k6ePCg6tSpo/j4eN15551au3at7rnnHp04cUIBAQGSpPnz5+v555/Xn3/+KXd3dz3//PNas2aNfvjhB/O1+vTpo5SUFMXExBSotrS0NPn5+Sk1NVW+vr7OH/wNChv1gdUlAIBT7Zra1+oSAADIlyPZ4KY+Q3a51NRUSVLp0qUlSbt27VJWVpY6dOhg9qldu7YqVaqk+Ph4SVJ8fLzq169vhjFJCg8PV1pamvbv32/2uXQbuX1yt5GfzMxMpaWl2T0AAAAAwBH/mkCWk5OjYcOGqUWLFqpXr54kKTExUe7u7vL397frGxAQoMTERLPPpWEsd3nusqv1SUtL07lz5/KtZ/LkyfLz8zMfFStWvOExAgAAAChe/jWBLCoqSj/88IM+/vhjq0uRJI0ZM0apqanm4/jx41aXBAAAAOBfpoTVBRTE4MGDtXr1am3evFkVKlQw2wMDA3XhwgWlpKTYnSVLSkpSYGCg2WfHjh1228udhfHSPpfPzJiUlCRfX195eXnlW5OHh4c8PDxueGwAAAAAiq+b+gyZYRgaPHiwVq5cqY0bNyokJMRueVhYmNzc3BQbG2u2HTp0SMeOHVOzZs0kSc2aNdO+ffuUnJxs9tmwYYN8fX0VGhpq9rl0G7l9crcBAAAAAIXhpj5DFhUVpaVLl+qLL76Qj4+Pec+Xn5+fvLy85OfnpwEDBmjEiBEqXbq0fH199cwzz6hZs2a68847JUmdOnVSaGioHnnkEUVHRysxMVEvvviioqKizDNcAwcO1OzZs/Xcc8/pscce08aNG/Xpp59qzZo1lo0dAAAAQNF3U58hmzdvnlJTU9W2bVsFBQWZj08++cTs88Ybb+iee+5Rr1691Lp1awUGBmrFihXmcldXV61evVqurq5q1qyZHn74YfXt21evvPKK2SckJERr1qzRhg0b1KBBA02fPl3vvvuuwsPD/9HxAgAAAChe/lWfQ3Yz43PIAOCfxeeQAQBuVkX2c8gAAAAAoCghkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZAAAAABgEQIZAAAAAFiEQAYAAAAAFiGQAQAAAIBFCGQAAAAAYBECGQAAAABYhEAGAAAAABYhkAEAAACARQhkAAAAAGARAhkAAAAAWIRABgAAAAAWIZABAAAAgEUIZJeZM2eOqlSpIk9PTzVt2lQ7duywuiQAAAAARRSB7BKffPKJRowYoZdfflnff/+9GjRooPDwcCUnJ1tdGgAAAIAiqITVBdxMZsyYoSeeeEL9+/eXJM2fP19r1qzR+++/r9GjR1tcHQAAN+bYK/WtLgEAnKrSuH1Wl3DDCGT/58KFC9q1a5fGjBljtrm4uKhDhw6Kj4/P0z8zM1OZmZnm89TUVElSWlpa4Rd7HbIzz1ldAgA41c36fnszO3M+2+oSAMCpbtbfBbl1GYZxzb4Esv/z119/KTs7WwEBAXbtAQEBOnjwYJ7+kydP1oQJE/K0V6xYsdBqBAD8f35vDbS6BACA1Sb7WV3BVZ05c0Z+flevkUB2ncaMGaMRI0aYz3NycnTq1CmVKVNGNpvNwsoA66SlpalixYo6fvy4fH19rS4HAGABfhcAf58ZO3PmjIKDg6/Zl0D2f2699Va5uroqKSnJrj0pKUmBgYF5+nt4eMjDw8Ouzd/fvzBLBP41fH19+SUMAMUcvwtQ3F3rzFguZln8P+7u7goLC1NsbKzZlpOTo9jYWDVr1szCygAAAAAUVZwhu8SIESPUr18/NW7cWE2aNNGbb76pjIwMc9ZFAAAAAHAmAtklHnjgAf35558aN26cEhMT1bBhQ8XExOSZ6ANA/jw8PPTyyy/nuZwXAFB88LsAcIzNKMhcjAAAAAAAp+MeMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAFimbdu2GjZsmNVlAAAcYLPZtGrVKqvLAIoMAhlQjM2fP18+Pj66ePGi2Zaeni43Nze1bdvWru+mTZtks9n0yy+//MNVAgD+SYmJiXrmmWdUtWpVeXh4qGLFirr33nvtPqsVgPMQyIBirF27dkpPT9d3331ntm3ZskWBgYFKSEjQ+fPnzfa4uDhVqlRJ1apVc+g1DMOwC3wAgJvX0aNHFRYWpo0bN2rq1Knat2+fYmJi1K5dO0VFRRXa6164cKHQtg3c7AhkQDFWq1YtBQUFadOmTWbbpk2b1K1bN4WEhOg///mPXXu7du2UmZmpIUOGqFy5cvL09FTLli21c+dOu342m01r165VWFiYPDw8tHXrVmVkZKhv374qVaqUgoKCNH369Dz1zJ07VzVq1JCnp6cCAgJ03333Fer4AQD2Bg0aJJvNph07dqhXr16qWbOm6tatqxEjRtj9Tvjrr7/Uo0cPlSxZUjVq1NCXX35pLsvOztaAAQMUEhIiLy8v1apVSzNnzrR7nUcffVTdu3fXpEmTFBwcrFq1akni9wCKJwIZUMy1a9dOcXFx5vO4uDi1bdtWbdq0MdvPnTunhIQEtWvXTs8995yWL1+uxYsX6/vvv1f16tUVHh6uU6dO2W139OjRev3113XgwAHddtttGjVqlL799lt98cUXWr9+vTZt2qTvv//e7P/dd99pyJAheuWVV3To0CHFxMSodevW/8xOAADo1KlTiomJUVRUlLy9vfMs9/f3N7+eMGGCevfurb1796pLly6KjIw0fw/k5OSoQoUK+uyzz/Tjjz9q3LhxeuGFF/Tpp5/abS82NlaHDh3Shg0btHr1an4PoPgyABRr77zzjuHt7W1kZWUZaWlpRokSJYzk5GRj6dKlRuvWrQ3DMIzY2FhDknH06FHDzc3NWLJkibn+hQsXjODgYCM6OtowDMOIi4szJBmrVq0y+5w5c8Zwd3c3Pv30U7Pt5MmThpeXlzF06FDDMAxj+fLlhq+vr5GWlvYPjBoAcLmEhARDkrFixYqr9pNkvPjii+bz9PR0Q5Kxdu3aK64TFRVl9OrVy3zer18/IyAgwMjMzDTb+D2A4oozZEAx17ZtW2VkZGjnzp3asmWLatasqbJly6pNmzbmfWSbNm1S1apVlZqaqqysLLVo0cJc383NTU2aNNGBAwfsttu4cWPz619++UUXLlxQ06ZNzbbSpUubl6hIUseOHVW5cmVVrVpVjzzyiJYsWaKzZ88W4sgBAJcyDKPAfW+77Tbza29vb/n6+io5OdlsmzNnjsLCwlS2bFmVKlVKCxYs0LFjx+y2Ub9+fbm7u5vP+T2A4opABhRz1atXV4UKFRQXF6e4uDi1adNGkhQcHKyKFStq+/btiouL01133eXQdvO73OVqfHx89P3332vZsmUKCgrSuHHj1KBBA6WkpDi0HQDA9alRo4ZsNpsOHjx4zb5ubm52z202m3JyciRJH3/8sZ599lkNGDBA69ev1549e9S/f/88E3dc/nuC3wMorghkANSuXTtt2rRJmzZtspvuvnXr1lq7dq127Nihdu3aqVq1anJ3d9e2bdvMPllZWdq5c6dCQ0OvuP1q1arJzc1NCQkJZtvp06f1008/2fUrUaKEOnTooOjoaO3du1dHjx7Vxo0bnTdQAMAVlS5dWuHh4ZozZ44yMjLyLC9oMNq2bZuaN2+uQYMG6fbbb1f16tUL/JEp/B5AcVTC6gIAWC93OuOsrCzzDJkktWnTRoMHD9aFCxfUrl07eXt76+mnn9aoUaNUunRpVapUSdHR0Tp79qwGDBhwxe2XKlVKAwYM0KhRo1SmTBmVK1dOY8eOlYvL//+f0OrVq/Xrr7+qdevWuuWWW/T1118rJyfH7rJGAEDhmjNnjlq0aKEmTZrolVde0W233aaLFy9qw4YNmjdvXp7L0/NTo0YNffDBB1q3bp1CQkL04YcfaufOnQoJCbnqevweQHFFIAOgdu3a6dy5c6pdu7YCAgLM9jZt2ujMmTPm9PiS9PrrrysnJ0ePPPKIzpw5o8aNG2vdunW65ZZbrvoaU6dOVXp6uu699175+Pho5MiRSk1NNZf7+/trxYoVGj9+vM6fP68aNWpo2bJlqlu3buEMGgCQR9WqVfX9999r0qRJGjlypP744w+VLVtWYWFhmjdvXoG28dRTT2n37t164IEHZLPZ9OCDD2rQoEFau3btVdfj9wCKK5vhyB2cAAAAAACn4R4yAAAAALAIgQwAAAAALEIgAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDABQpCUmJuqZZ55R1apV5eHhoYoVK+ree+9VbGxsgdZftGiR/P39C7dIAECxVcLqAgAAKCxHjx5VixYt5O/vr6lTp6p+/frKysrSunXrFBUVpYMHD1pdosOysrLk5uZmdRkAACfhDBkAoMgaNGiQbDabduzYoV69eqlmzZqqW7euRowYof/85z+SpBkzZqh+/fry9vZWxYoVNWjQIKWnp0uSNm3apP79+ys1NVU2m002m03jx4+XJGVmZurZZ59V+fLl5e3traZNm2rTpk12r//OO++oYsWKKlmypHr06KEZM2bkOds2b948VatWTe7u7qpVq5Y+/PBDu+U2m03z5s1T165d5e3trYkTJ6p69eqaNm2aXb89e/bIZrPp559/dt4OBAAUOgIZAKBIOnXqlGJiYhQVFSVvb+88y3ODkYuLi2bNmqX9+/dr8eLF2rhxo5577jlJUvPmzfXmm2/K19dXf/zxh/744w89++yzkqTBgwcrPj5eH3/8sfbu3av7779fnTt31uHDhyVJ27Zt08CBAzV06FDt2bNHHTt21KRJk+xqWLlypYYOHaqRI0fqhx9+0FNPPaX+/fsrLi7Ort/48ePVo0cP7du3TwMGDNBjjz2mhQsX2vVZuHChWrdurerVqztl/wEA/hk2wzAMq4sAAMDZduzYoaZNm2rFihXq0aNHgdf7/PPPNXDgQP3111+S/r6HbNiwYUpJSTH7HDt2TFWrVtWxY8cUHBxstnfo0EFNmjTRa6+9pj59+ig9PV2rV682lz/88MNavXq1ua0WLVqobt26WrBggdmnd+/eysjI0Jo1ayT9fYZs2LBheuONN8w+J06cUKVKlbR9+3Y1adJEWVlZCg4O1rRp09SvXz+H9hMAwFqcIQMAFEkF/X/jN998o/bt26t8+fLy8fHRI488opMnT+rs2bNXXGffvn3Kzs5WzZo1VapUKfPx7bff6pdffpEkHTp0SE2aNLFb7/LnBw4cUIsWLezaWrRooQMHDti1NW7c2O55cHCwIiIi9P7770uSvvrqK2VmZur+++8v0JgBADcPJvUAABRJNWrUkM1mu+rEHUePHtU999yjp59+WpMmTVLp0qW1detWDRgwQBcuXFDJkiXzXS89PV2urq7atWuXXF1d7ZaVKlXKqeOQlO8ll48//rgeeeQRvfHGG1q4cKEeeOCBK9YLALh5cYYMAFAklS5dWuHh4ZozZ44yMjLyLE9JSdGuXbuUk5Oj6dOn684771TNmjV14sQJu37u7u7Kzs62a7v99tuVnZ2t5ORkVa9e3e4RGBgoSapVq5Z27txpt97lz+vUqaNt27bZtW3btk2hoaHXHF+XLl3k7e2tefPmKSYmRo899tg11wEA3HwIZACAImvOnDnKzs5WkyZNtHz5ch0+fFgHDhzQrFmz1KxZM1WvXl1ZWVl666239Ouvv+rDDz/U/Pnz7bZRpUoVpaenKzY2Vn/99ZfOnj2rmjVrKjIyUn379tWKFSt05MgR7dixQ5MnTzbv/XrmmWf09ddfa8aMGTp8+LDefvttrV27Vjabzdz2qFGjtGjRIs2bN0+HDx/WjBkztGLFCnPikKtxdXXVo48+qjFjxqhGjRpq1qyZc3ceAOCfYQAAUISdOHHCiIqKMipXrmy4u7sb5cuXN7p27WrExcUZhmEYM2bMMIKCggwvLy8jPDzc+OCDDwxJxunTp81tDBw40ChTpowhyXj55ZcNwzCMCxcuGOPGjTOqVKliuLm5GUFBQUaPHj2MvXv3mustWLDAKF++vOHl5WV0797dmDhxohEYGGhX39y5c42qVasabm5uRs2aNY0PPvjAbrkkY+XKlfmO7ZdffjEkGdHR0Te8nwAA1mCWRQAA/iFPPPGEDh48qC1btjhle1u2bFH79u11/PhxBQQEOGWbAIB/FpN6AABQSKZNm6aOHTvK29tba9eu1eLFizV37twb3m5mZqb+/PNPjR8/Xvfffz9hDAD+xbiHDACAQrJjxw517NhR9evX1/z58zVr1iw9/vjjN7zdZcuWqXLlykpJSVF0dLQTKgUAWIVLFgEAAADAIpwhAwAAAACLEMgAAAAAwCIEMgAAAACwCIEMAAAAACxCIAMAAAAAixDIAAAAAMAiBDIAAAAAsAiBDAAAAAAs8v8AHgJIagAUDJ4AAAAASUVORK5CYII="},"metadata":{}}]},{"cell_type":"markdown","source":"## Data preparation for the model","metadata":{}},{"cell_type":"code","source":"# CHARS ENCODER DECODER\n\nchar_to_int = { char:i for i,char in enumerate(chars_from_text) }\nint_to_char = { i:char for i,char in enumerate(chars_from_text) }\nencode_chars = lambda char_sequence: [char_to_int[char] for char in char_sequence if char in char_to_int]\ndecode_chars = lambda int_sequence: ''.join([int_to_char[i] for i in int_sequence if i in int_to_char])\n\n# WORDS ENCODER DECODER\n\n# Encoder and decoder\n# These are responsible for turning our tokens from a text format to a number format\n# so that the model can use them for predicting. Remember that computers cannot understand\n# text, only numbers!\nword_to_int = { word:i for i,word in enumerate(words_from_text) }\nint_to_word = { i:word for i,word in enumerate(words_from_text) }\nencode_words = lambda word_sequence: [word_to_int[word.lower()] for word in word_sequence.split(' ') if word.lower() in word_to_int]\ndecode_words = lambda int_sequence: ' '.join([int_to_word[i] for i in int_sequence if i in int_to_word])\n\n# This is our whole data(text from the book) that we're going to use to\n# train and evaluate the model. The data is encoded(all the tokens are transformed\n# into a number format).\n# For example:\n# a -> 1\n# b -> 2\n# c -> 3\n# or if our tokens we're words\n# cat -> 1\n# dog -> 2\n# have -> 3\nchars_data = torch.tensor(encode_chars(text), dtype=torch.long)\nwords_data = torch.tensor(encode_words(text), dtype=torch.long)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:08.444657Z","iopub.execute_input":"2024-09-30T13:05:08.445043Z","iopub.status.idle":"2024-09-30T13:05:08.613004Z","shell.execute_reply.started":"2024-09-30T13:05:08.445014Z","shell.execute_reply":"2024-09-30T13:05:08.612280Z"},"trusted":true},"execution_count":5,"outputs":[]},{"cell_type":"code","source":"# The function for splitting data for training and validation\ndef train_val_split(data, split_rate=0.8):\n    split = int(split_rate * len(data))\n    train_data = data[:split]\n    val_data = data[split:]\n\n    return train_data, val_data","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:10.655803Z","iopub.execute_input":"2024-09-30T13:05:10.656425Z","iopub.status.idle":"2024-09-30T13:05:10.661252Z","shell.execute_reply.started":"2024-09-30T13:05:10.656392Z","shell.execute_reply":"2024-09-30T13:05:10.660325Z"},"trusted":true},"execution_count":6,"outputs":[]},{"cell_type":"code","source":"chars_train_data, chars_val_data = train_val_split(chars_data)\nwords_train_data, words_val_data = train_val_split(words_data)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:12.500006Z","iopub.execute_input":"2024-09-30T13:05:12.500357Z","iopub.status.idle":"2024-09-30T13:05:12.515452Z","shell.execute_reply.started":"2024-09-30T13:05:12.500328Z","shell.execute_reply":"2024-09-30T13:05:12.514509Z"},"trusted":true},"execution_count":7,"outputs":[]},{"cell_type":"code","source":"# A function to split data into batches so we don't\n# give the model everything while training it, but\n# chunks of data by chunks of data.\ndef get_batch(train_data, val_data, data_type='train'):\n    data = train_data if data_type == 'train' else val_data\n    rand_batch_selection = torch.randint(len(data) - block_size, (batch_size,))\n    x = torch.stack([data[i:i+batch_size] for i in rand_batch_selection])\n    y = torch.stack([data[i+1:i+batch_size+1] for i in rand_batch_selection])\n    x.to(device), y.to(device)\n    return (x,y)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:14.220764Z","iopub.execute_input":"2024-09-30T13:05:14.221415Z","iopub.status.idle":"2024-09-30T13:05:14.227496Z","shell.execute_reply.started":"2024-09-30T13:05:14.221383Z","shell.execute_reply":"2024-09-30T13:05:14.226510Z"},"trusted":true},"execution_count":8,"outputs":[]},{"cell_type":"markdown","source":"## The model variants","metadata":{}},{"cell_type":"code","source":"class BigramLanguageModel(nn.Module):\n    def __init__(self, vocab_size):\n        super().__init__()\n        # We make an embedding table to know how related\n        # are the words/chars to each other. Which one we should place next.\n        self.vocab_size = vocab_size\n        self.token_embedding_table = nn.Embedding(vocab_size, vocab_size, device=device)\n        \n    # Forward propagation of our input throught the model to see what we get as output.\n    # These results later get used for backwards propagation to update the weights of the model.\n    def forward(self, index, targets=None):\n        index = index.to(self.token_embedding_table.weight.device)\n        \n        probs = self.token_embedding_table(index)\n        \n        if targets == None:\n            # If we don't have the targets, we don't have a reference on which to calculate the loss\n            loss = None\n        else:\n            # Make sure the targets are on the same device as the embedding table (cpu/gpu)\n            targets = targets.to(self.token_embedding_table.weight.device)\n            \n            # We do this reshaping of the probabilities because torch has certain rules on\n            # how it expects its data shape to be\n            B, T, C = probs.shape # Batch - how many context examples do we give the model, Time - tokens(what is the next char/word), Channels - vocabulary size\n            probs = probs.view(B*T, C)\n            targets = targets.view(B*T)\n            # We calculate the differences between our models probabilities and what they should actually be\n            # to know what is our loss and how well the model is performing.\n            loss = F.cross_entropy(probs, targets)\n        \n        return probs, loss\n        \n    # Function to generate the next token in the sequence based on the context it was provided.\n    def generate(self, idx, generate_max_tokens):\n        for _ in range(generate_max_tokens):\n            idx = idx.to(self.token_embedding_table.weight.device)\n            # Forward propagation to obtain probabilities and the loss.\n            probs, loss = self(idx)\n            # Take the last tokens probabilities for the next token to be generated in the sequence.\n            probs = probs[:, -1, :]\n            # We apply the softmax function to normalize and \"smooth\" the selection.\n            probs = F.softmax(probs, dim=-1)\n            # Pick the most probable token in the sequence with the multinomial distribution.\n            idx_next = torch.multinomial(probs, num_samples=1)\n            # And at the end concatenate it to the previous context.\n            idx = torch.cat((idx, idx_next), dim=1)\n        return idx\n    \n    def refresh_weights(self):\n        self.token_embedding_table = nn.Embedding(self.vocab_size, self.vocab_size, device=device)\n               \n            \nmodel_chars = BigramLanguageModel(vocab_size_chars)\n# Make sure the model is on the same device as the rest of the data.\nmcb = model_chars.to(device)\nmodel_words = BigramLanguageModel(vocab_size_words)\nmwb = model_words.to(device)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:15.837029Z","iopub.execute_input":"2024-09-30T13:05:15.837351Z","iopub.status.idle":"2024-09-30T13:05:16.035667Z","shell.execute_reply.started":"2024-09-30T13:05:15.837324Z","shell.execute_reply":"2024-09-30T13:05:16.034898Z"},"trusted":true},"execution_count":9,"outputs":[]},{"cell_type":"code","source":"class Head(nn.Module):\n    def __init__(self, head_size):\n        super().__init__()\n        self.key = nn.Linear(n_embd, head_size, bias=False)\n        self.query = nn.Linear(n_embd, head_size, bias=False)\n        self.value = nn.Linear(n_embd, head_size, bias=False)\n        self.register_buffer('tril', torch.tril(torch.ones(block_size, block_size)))\n        \n        self.dropout = nn.Dropout(dropout)\n        \n    def forward(self, x):\n        B, T, C = x.shape\n        # What I have.\n        k = self.key(x)\n        # What I'm looking for.\n        q = self.query(x)\n        # Calculate attention scores and normalize them by n_embd.\n        wei = q @ k.transpose(-1, -2) * C**-0.5\n        # Don't communicate with the future only the past for predictions.\n        wei = wei.masked_fill(self.tril[:T,:T] == 0, float('-inf'))\n        wei = F.softmax(wei, dim=-1)\n        wei = self.dropout(wei)\n        v = self.value(x)\n        out = wei @ v\n        return out\n    \nclass MultiHeadAttention(nn.Module):\n    def __init__(self, num_heads, head_size):\n        super().__init__()\n        self.heads = nn.ModuleList([Head(head_size) for _ in range(num_heads)])\n        self.proj = nn.Linear(n_embd, n_embd)\n        self.dropout = nn.Dropout(dropout)\n        \n    def forward(self, x):\n        out = torch.cat([h(x) for h in self.heads], dim=-1)\n        out = self.dropout(self.proj(out))\n        return out\n        \nclass FeedForward(nn.Module):\n    def __init__(self, n_embd):\n        super().__init__()\n        self.net = nn.Sequential(\n            nn.Linear(n_embd, 4 * n_embd),\n            nn.ReLU(),\n            nn.Linear(4 * n_embd, n_embd),\n            nn.Dropout(dropout),\n        )\n    \n    def forward(self, x):\n        return self.net(x)\n    \nclass Block(nn.Module):\n    def __init__(self, n_embd, n_head):\n        super().__init__()\n        head_size = n_embd // n_head\n        self.sa = MultiHeadAttention(n_head, head_size)\n        self.ffwd = FeedForward(n_embd)\n        self.ln1 = nn.LayerNorm(n_embd)\n        self.ln2 = nn.LayerNorm(n_embd)\n    \n    def forward(self, x):\n        y = self.sa(x)\n        x = self.ln1(x + y)\n        y = self.ffwd(x)\n        x = self.ln2(x + y)\n        return x\n\nclass GPTLanguageModel(nn.Module):\n    def __init__(self, vocab_size):\n        super().__init__()\n        # We make an embedding table to know how related\n        # are the words/chars to each other. Which one we should place next.\n        self.vocab_size = vocab_size\n        # Regular token embeddings.\n        self.token_embedding_table = nn.Embedding(vocab_size, n_embd,device=device)\n        # Token positional embeddings.\n        self.position_embedding_table = nn.Embedding(block_size, n_embd, device=device)\n        self.blocks = nn.Sequential(*[Block(n_embd, n_head=n_embd//4) for _ in range(n_layers)])\n        self.lm_head = nn.Linear(n_embd, vocab_size)\n        \n    # Forward propagation of our input throught the model to see what we get as output.\n    # These results later get used for backwards propagation to update the weights of the model.\n    def forward(self, index, targets=None):\n        index = index.to(self.token_embedding_table.weight.device)\n        B, T = index.shape\n        \n        # Token embeddings\n        # What token should go in this position based on previous context(tokens)?\n        probs_tok = self.token_embedding_table(index)\n        # Position of the token embeddings\n        # What token has affinity to be in this position?\n        probs_pos = self.position_embedding_table(torch.arange(T, device=device))\n        probs = probs_tok + probs_pos\n        probs = self.blocks(probs)\n        # Decoder, output probs for number of vocab_size tokens we have.\n        probs = self.lm_head(probs)\n        \n        if targets == None:\n            # If we don't have the targets, we don't have a reference on which to calculate the loss\n            loss = None\n        else:\n            # Make sure the targets are on the same device as the embedding table (cpu/gpu)\n            targets = targets.to(self.token_embedding_table.weight.device)\n            \n            # We do this reshaping of the probabilities because torch has certain rules on\n            # how it expects its data shape to be\n            B, T, C = probs.shape # Batch - how many context examples do we give the model, Time - tokens(what is the next char/word), Channels - vocabulary size\n            probs = probs.view(B*T, C)\n            targets = targets.view(B*T)\n            # We calculate the differences between our models probabilities and what they should actually be\n            # to know what is our loss and how well the model is performing.\n            loss = F.cross_entropy(probs, targets)\n        \n        return probs, loss\n        \n    # Function to generate the next token in the sequence based on the context it was provided.\n    def generate(self, idx, generate_max_tokens):\n        for _ in range(generate_max_tokens):\n            idx = idx.to(self.token_embedding_table.weight.device)\n            # Crop last to last block_size tokens, because this is our max context,\n            # else we will get problem in forward function with positional embedding.\n            idx_context = idx[:, -block_size:]\n            # Forward propagation to obtain probabilities and the loss.\n            probs, loss = self(idx_context)\n            # Take the last tokens probabilities for the next token to be generated in the sequence.\n            probs = probs[:, -1, :]\n            # We apply the softmax function to normalize and \"smooth\" the selection.\n            probs = F.softmax(probs, dim=-1)\n            # Pick the most probable token in the sequence with the multinomial distribution.\n            idx_next = torch.multinomial(probs, num_samples=1)\n            # And at the end concatenate it to the previous context.\n            idx = torch.cat((idx, idx_next), dim=1)\n        return idx\n    \n    def refresh_weights(self):\n        self.token_embedding_table = nn.Embedding(self.vocab_size, n_embd, device=device)\n        self.position_embedding_table = nn.Embedding(block_size, n_embd, device=device)\n               \n            \nmodel_chars = GPTLanguageModel(vocab_size_chars)\n# Make sure the model is on the same device as the rest of the data.\nmc = model_chars.to(device)\nmodel_words = GPTLanguageModel(vocab_size_words)\nmw = model_words.to(device)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:17.534573Z","iopub.execute_input":"2024-09-30T13:05:17.534888Z","iopub.status.idle":"2024-09-30T13:05:19.016279Z","shell.execute_reply.started":"2024-09-30T13:05:17.534863Z","shell.execute_reply":"2024-09-30T13:05:19.015484Z"},"trusted":true},"execution_count":10,"outputs":[]},{"cell_type":"code","source":"# The optimizer which will be used to update the weights and biases of the model\n# during back propagation.\n# GPT model optimizers\noptimizer_chars = torch.optim.AdamW(mc.parameters(), lr=3e-4)\noptimizer_words = torch.optim.AdamW(mw.parameters(), lr=3e-4)\n# Bigram model optimizers\noptimizer_chars_bigram = torch.optim.AdamW(mcb.parameters(), lr=1e-2)\noptimizer_words_bigram = torch.optim.AdamW(mwb.parameters(), lr=1e-2)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:22.864241Z","iopub.execute_input":"2024-09-30T13:05:22.864963Z","iopub.status.idle":"2024-09-30T13:05:24.022828Z","shell.execute_reply.started":"2024-09-30T13:05:22.864932Z","shell.execute_reply":"2024-09-30T13:05:24.021957Z"},"trusted":true},"execution_count":11,"outputs":[]},{"cell_type":"code","source":"# Model type can be either 'chars' or 'words'.\n# If bigram_gpt is False gpt will be used, else bigram.\ndef train(model_type, bigram_gpt=False, steps=1000, loss_freq_tracking = 100):\n    # Evaluate loss every loss_freq_tracking steps.\n    losses = torch.zeros(steps//loss_freq_tracking)\n    for step in range(steps):\n        x, y = None, None\n        probs, loss = None, None\n        # For char models.\n        if model_type == 'chars':\n            x, y = get_batch(chars_train_data, chars_val_data)\n            if bigram_gpt:\n                probs, loss = mcb(x, y)\n                optimizer_chars_bigram.zero_grad(set_to_none=True)\n                loss.backward()\n                optimizer_chars_bigram.step()\n            else:\n                probs, loss = mc(x, y)\n                optimizer_chars.zero_grad(set_to_none=True)\n                loss.backward()\n                optimizer_chars.step()\n        # For word models.\n        else:\n            x, y = get_batch(words_train_data, words_val_data)\n            if bigram_gpt:\n                probs, loss = mwb(x, y)\n                optimizer_words_bigram.zero_grad(set_to_none=True)\n                loss.backward()\n                optimizer_words_bigram.step()\n            else:\n                probs, loss = mw(x, y)\n                optimizer_words.zero_grad(set_to_none=True)\n                loss.backward()\n                optimizer_words.step()\n        if step % loss_freq_tracking:\n            losses[step//loss_freq_tracking] = loss.item()\n    \n    # Plot for the loss during training.\n    plt.plot(losses, label=\"loss\")\n    plt.title(\"Loss during training\")\n    plt.ylabel(\"Loss\")\n    plt.xlabel(\"Step x %s\" % loss_freq_tracking)\n    plt.legend()\n    plt.show()\n\n    print(\"Current loss: \", loss.item())","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:27.534173Z","iopub.execute_input":"2024-09-30T13:05:27.534672Z","iopub.status.idle":"2024-09-30T13:05:27.545537Z","shell.execute_reply.started":"2024-09-30T13:05:27.534643Z","shell.execute_reply":"2024-09-30T13:05:27.544484Z"},"trusted":true},"execution_count":12,"outputs":[]},{"cell_type":"markdown","source":"Training of the Bigram char model","metadata":{}},{"cell_type":"code","source":"train('chars', True, steps=1000, loss_freq_tracking=100)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:50:10.929292Z","iopub.execute_input":"2024-09-14T15:50:10.930024Z","iopub.status.idle":"2024-09-14T15:50:14.315187Z","shell.execute_reply.started":"2024-09-14T15:50:10.929990Z","shell.execute_reply":"2024-09-14T15:50:14.314207Z"},"trusted":true},"execution_count":69,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"name":"stdout","text":"Current loss:  4.093766689300537\n","output_type":"stream"}]},{"cell_type":"code","source":"mcb.refresh_weights()","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:49:57.217727Z","iopub.execute_input":"2024-09-14T15:49:57.218089Z","iopub.status.idle":"2024-09-14T15:49:57.223091Z","shell.execute_reply.started":"2024-09-14T15:49:57.218056Z","shell.execute_reply":"2024-09-14T15:49:57.222124Z"},"trusted":true},"execution_count":66,"outputs":[]},{"cell_type":"markdown","source":"Training of the GPT char model","metadata":{}},{"cell_type":"code","source":"train('chars', steps=1000, loss_freq_tracking=100)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:30:48.277918Z","iopub.execute_input":"2024-09-14T15:30:48.278629Z","iopub.status.idle":"2024-09-14T15:41:35.765147Z","shell.execute_reply.started":"2024-09-14T15:30:48.278596Z","shell.execute_reply":"2024-09-14T15:41:35.764201Z"},"trusted":true},"execution_count":52,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"name":"stdout","text":"Current loss:  1.5991235971450806\n","output_type":"stream"}]},{"cell_type":"code","source":"mc.refresh_weights()","metadata":{"execution":{"iopub.status.busy":"2024-09-07T11:34:15.522456Z","iopub.execute_input":"2024-09-07T11:34:15.523303Z","iopub.status.idle":"2024-09-07T11:34:15.527993Z","shell.execute_reply.started":"2024-09-07T11:34:15.523267Z","shell.execute_reply":"2024-09-07T11:34:15.527137Z"},"trusted":true},"execution_count":51,"outputs":[]},{"cell_type":"markdown","source":"Training of the Bigram word model","metadata":{}},{"cell_type":"code","source":"train('words', True, steps=1000,  loss_freq_tracking=100)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:29:07.658404Z","iopub.execute_input":"2024-09-14T15:29:07.659271Z","iopub.status.idle":"2024-09-14T15:29:59.329628Z","shell.execute_reply.started":"2024-09-14T15:29:07.659228Z","shell.execute_reply":"2024-09-14T15:29:59.328726Z"},"trusted":true},"execution_count":51,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"name":"stdout","text":"Current loss:  3.7079832553863525\n","output_type":"stream"}]},{"cell_type":"markdown","source":"Training of the GPT word model","metadata":{}},{"cell_type":"code","source":"mwb.refresh_weights()","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:27:49.197086Z","iopub.execute_input":"2024-09-14T15:27:49.197938Z","iopub.status.idle":"2024-09-14T15:27:49.202466Z","shell.execute_reply.started":"2024-09-14T15:27:49.197905Z","shell.execute_reply":"2024-09-14T15:27:49.201428Z"},"trusted":true},"execution_count":46,"outputs":[]},{"cell_type":"code","source":"train('words', False, steps=1000, loss_freq_tracking = 100)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:02:33.151823Z","iopub.execute_input":"2024-09-14T15:02:33.152625Z","iopub.status.idle":"2024-09-14T15:03:40.475168Z","shell.execute_reply.started":"2024-09-14T15:02:33.152589Z","shell.execute_reply":"2024-09-14T15:03:40.474268Z"},"trusted":true},"execution_count":20,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"name":"stdout","text":"Current loss:  6.4558820724487305\n","output_type":"stream"}]},{"cell_type":"code","source":"mw.refresh_weights()","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:28:56.730217Z","iopub.execute_input":"2024-09-14T15:28:56.730496Z","iopub.status.idle":"2024-09-14T15:28:56.744808Z","shell.execute_reply.started":"2024-09-14T15:28:56.730474Z","shell.execute_reply":"2024-09-14T15:28:56.743953Z"},"trusted":true},"execution_count":50,"outputs":[]},{"cell_type":"markdown","source":"This function is to run and see bahaviour of different parameters so you can tune your model the best. It might take some time to execute.","metadata":{}},{"cell_type":"code","source":"# Model type can be either 'chars' or 'words'.\n# If bigram_gpt is False gpt will be used, else bigram.\ndef eval_model_with_movable_parameters(model_type, bigram_gpt=False):\n    for lr in (1e-2, 1e-3,1e-4, 1e-5, 1e-6, 1e-7, 1e-8):\n        print(\"Learning rate: \", lr)\n        if model_type == 'chars':\n            pass\n            optimizer_chars = torch.optim.AdamW(mc.parameters(), lr=lr)\n        else:\n            optimizer_words = torch.optim.AdamW(mw.parameters(), lr=lr)\n        \n        for block_size in (4, 6, 8, 10, 12):\n            print(\"Block size: \", block_size)\n            if model_type == 'chars':\n                if bigram_gpt:\n                    mcb.refresh_weights()\n                    train(model_type, True, steps=1000, loss_freq_tracking=100)\n                else:\n                    mc.refresh_weights()\n                    train(model_type, steps=1000, loss_freq_tracking=100)\n            else:\n                if bigram_gpt:\n                    mwb.refresh_weights()\n                    train(model_type, True, steps=1000, loss_freq_tracking=100)\n                else:\n                    mw.refresh_weights()\n                    train(model_type, steps=1000, loss_freq_tracking=100)","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:39.352804Z","iopub.execute_input":"2024-09-30T13:05:39.353271Z","iopub.status.idle":"2024-09-30T13:05:39.365509Z","shell.execute_reply.started":"2024-09-30T13:05:39.353233Z","shell.execute_reply":"2024-09-30T13:05:39.364488Z"},"trusted":true},"execution_count":13,"outputs":[]},{"cell_type":"code","source":"eval_model_with_movable_parameters('chars', True)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"eval_model_with_movable_parameters('words', True)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Model type can be either 'chars' or 'words'.\n# If bigram_gpt is False gpt will be used, else bigram.\ndef generate_prompt(model_type, bigram_gpt=False , generate_max_tokens=100):\n    model = None\n    if model_type == 'chars':\n        if bigram_gpt:\n            model = mcb\n        else:\n            model = mc\n        print((decode_chars(model.generate(torch.zeros((1,1), dtype=torch.long, device=device), generate_max_tokens)[0].tolist())))\n    else:\n        if bigram_gpt:\n            model = mwb\n        else:\n            model = mw\n        print((decode_words(model.generate(torch.zeros((1,1), dtype=torch.long, device=device), generate_max_tokens)[0].tolist())))","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:05:44.383570Z","iopub.execute_input":"2024-09-30T13:05:44.384183Z","iopub.status.idle":"2024-09-30T13:05:44.390831Z","shell.execute_reply.started":"2024-09-30T13:05:44.384149Z","shell.execute_reply":"2024-09-30T13:05:44.390067Z"},"trusted":true},"execution_count":14,"outputs":[]},{"cell_type":"markdown","source":"### Bigram Chars model","metadata":{}},{"cell_type":"code","source":"generate_prompt('chars', True)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:43:20.667986Z","iopub.execute_input":"2024-09-14T15:43:20.668956Z","iopub.status.idle":"2024-09-14T15:43:20.705859Z","shell.execute_reply.started":"2024-09-14T15:43:20.668909Z","shell.execute_reply":"2024-09-14T15:43:20.705009Z"},"trusted":true},"execution_count":54,"outputs":[{"name":"stdout","text":" g. e nghotome th! s ns. medse so blis ererisexushil athind thatharuf astheaithait ntom n t t wo sppo\n","output_type":"stream"}]},{"cell_type":"markdown","source":"### Transformer Chars model","metadata":{}},{"cell_type":"code","source":"generate_prompt('chars')","metadata":{"execution":{"iopub.status.busy":"2024-09-26T13:18:17.389856Z","iopub.execute_input":"2024-09-26T13:18:17.390829Z","iopub.status.idle":"2024-09-26T13:18:42.948603Z","shell.execute_reply.started":"2024-09-26T13:18:17.390795Z","shell.execute_reply":"2024-09-26T13:18:42.947831Z"},"trusted":true},"execution_count":18,"outputs":[{"name":"stdout","text":" unto shall all mydeheart not to long belong unto me  am n. hough wh by lingrevind thus t twhan is pr\n","output_type":"stream"}]},{"cell_type":"markdown","source":"### Bigram Words model","metadata":{}},{"cell_type":"code","source":"generate_prompt('words', True)","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:23:15.033928Z","iopub.execute_input":"2024-09-14T15:23:15.034431Z","iopub.status.idle":"2024-09-14T15:23:15.072158Z","shell.execute_reply.started":"2024-09-14T15:23:15.034383Z","shell.execute_reply":"2024-09-14T15:23:15.071347Z"},"trusted":true},"execution_count":40,"outputs":[{"name":"stdout","text":"\"(and officious man, it is not thou set in the some boasting, hordes from care, ape. \"new æther, spirit, and that they were \"honour either: as one, of in general ruler moderately, law: whereby and beauty at use of the troubles might be that they may not to set upon him contentedly, as souls were in some kind and cautelous, marcomanni or with thyself. for the common good, what crosses arose a god, or affected: no repent of his sanctity; black or clamour. moreover i am i bread polemon; physicians. for one of corn;' and effectual than dispersed, and such a\n","output_type":"stream"}]},{"cell_type":"markdown","source":"### Transformer Words model","metadata":{}},{"cell_type":"code","source":"generate_prompt('words')","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:07:14.021728Z","iopub.execute_input":"2024-09-30T13:07:14.022101Z","iopub.status.idle":"2024-09-30T13:07:40.609894Z","shell.execute_reply.started":"2024-09-30T13:07:14.022071Z","shell.execute_reply":"2024-09-30T13:07:40.608964Z"},"trusted":true},"execution_count":16,"outputs":[{"name":"stdout","text":"\"(and fondly imagine the cadiciant's, fabius, julianus lepidus, or any preconceived plan. he into drunkenness and it is divided and separated; or so affixed, so congealed and as it were that is last breath them, is in no sure and countenance they that shall we have no to do the earth doth end, and bestow so intend the one swarms, and another ground. the them, yet hast taken i am bound with so many ways, better and their grief of them be a sudden that some one or as it away. i consist freely to be a perpetual grief and pass\n","output_type":"stream"}]},{"cell_type":"code","source":"# Function for average estimation of loss.\n# It is used to get a better view on what the loss is.\n# Model type can be either 'chars' or 'words'.\n# If bigram_gpt is False gpt will be used, else bigram.\ndef estimate_loss(model_type, bigram_gpt=False):\n    out = {}\n    # Turn on validation mode.\n    if model_type == 'chars':\n        mc.eval()\n    else:\n        mw.eval()\n    # We turn off gradient calculation here\n    # because we're just evaluating the model not training it.\n    # (We aren't updating the weights!!!)\n    with torch.no_grad():\n        for data_type in ['train', 'val']:\n            losses = torch.zeros(eval_iters)\n            for i in range(eval_iters):\n                x, y = None, None\n                probs, loss = None, None\n                if model_type == 'chars':\n                    x, y = get_batch(chars_train_data, chars_val_data, data_type)\n                    # Could be also called logits.\n                    # Logits are unnormalized probabilities!\n                    if bigram_gpt:\n                        probs, loss = mcb(x, y)\n                    else:\n                        probs, loss = mc(x, y)\n                else:\n                    x, y = get_batch(words_train_data, words_val_data, data_type)\n                    # Could be also called logits.\n                    # Logits are unnormalized probabilities!\n                    probs, loss = mw(x, y)\n                # Gives us just the loss value from tensor.\n                losses[i] = loss.item()\n            # Add data to the plot so it can be displayed.\n            plt.plot(losses, label=data_type)\n            out[data_type] = losses.mean()\n    plt.title(\"Model loss estimation\")\n    plt.ylabel(\"Loss\")\n    plt.xlabel(\"Eval iteration\")\n    plt.legend()\n    plt.show()\n    # Turn the training mode after the eval mode.\n    if model_type == 'chars':\n        mc.train()\n    else:\n        mw.train()\n    return out","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:01:53.024979Z","iopub.execute_input":"2024-09-14T15:01:53.025747Z","iopub.status.idle":"2024-09-14T15:01:53.035656Z","shell.execute_reply.started":"2024-09-14T15:01:53.025716Z","shell.execute_reply":"2024-09-14T15:01:53.034751Z"},"trusted":true},"execution_count":17,"outputs":[]},{"cell_type":"markdown","source":"Estimation of the Bigram char model","metadata":{}},{"cell_type":"code","source":"estimate_loss('chars', True)","metadata":{"execution":{"iopub.status.busy":"2024-08-30T17:21:04.200496Z","iopub.execute_input":"2024-08-30T17:21:04.200876Z","iopub.status.idle":"2024-08-30T17:21:04.577173Z","shell.execute_reply.started":"2024-08-30T17:21:04.200843Z","shell.execute_reply":"2024-08-30T17:21:04.576196Z"},"trusted":true},"execution_count":247,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"execution_count":247,"output_type":"execute_result","data":{"text/plain":"{'train': tensor(2.6343), 'val': tensor(2.8581)}"},"metadata":{}}]},{"cell_type":"markdown","source":"Estimation of the GPT char model","metadata":{}},{"cell_type":"code","source":"estimate_loss('chars')","metadata":{"execution":{"iopub.status.busy":"2024-09-07T13:31:28.226055Z","iopub.execute_input":"2024-09-07T13:31:28.226894Z","iopub.status.idle":"2024-09-07T13:31:31.984541Z","shell.execute_reply.started":"2024-09-07T13:31:28.226860Z","shell.execute_reply":"2024-09-07T13:31:31.983656Z"},"trusted":true},"execution_count":75,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"execution_count":75,"output_type":"execute_result","data":{"text/plain":"{'train': tensor(0.5821), 'val': tensor(2.1705)}"},"metadata":{}}]},{"cell_type":"markdown","source":"Estimation of the Bigram word model","metadata":{}},{"cell_type":"code","source":"estimate_loss('words', True)","metadata":{"execution":{"iopub.status.busy":"2024-08-31T10:51:02.848051Z","iopub.execute_input":"2024-08-31T10:51:02.848780Z","iopub.status.idle":"2024-08-31T10:51:06.983318Z","shell.execute_reply.started":"2024-08-31T10:51:02.848739Z","shell.execute_reply":"2024-08-31T10:51:06.982447Z"},"trusted":true},"execution_count":193,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"execution_count":193,"output_type":"execute_result","data":{"text/plain":"{'train': tensor(9.2117), 'val': tensor(9.2259)}"},"metadata":{}}]},{"cell_type":"markdown","source":"Estimation of the GPT word model","metadata":{}},{"cell_type":"code","source":"estimate_loss('words')","metadata":{"execution":{"iopub.status.busy":"2024-09-14T15:04:29.892333Z","iopub.execute_input":"2024-09-14T15:04:29.893055Z","iopub.status.idle":"2024-09-14T15:04:33.803184Z","shell.execute_reply.started":"2024-09-14T15:04:29.893025Z","shell.execute_reply":"2024-09-14T15:04:33.802317Z"},"trusted":true},"execution_count":21,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 1 Axes>","image/png":"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"},"metadata":{}},{"execution_count":21,"output_type":"execute_result","data":{"text/plain":"{'train': tensor(6.3526), 'val': tensor(7.9203)}"},"metadata":{}}]},{"cell_type":"markdown","source":"## Weights exportation for deployment","metadata":{}},{"cell_type":"markdown","source":"## Bigram","metadata":{}},{"cell_type":"markdown","source":"chars","metadata":{}},{"cell_type":"code","source":"# Weights export for chars Bigram\ntorch.save(mcb.state_dict(), 'model_chars_weights_bigram.pth')","metadata":{"execution":{"iopub.status.busy":"2024-09-07T14:53:54.791470Z","iopub.execute_input":"2024-09-07T14:53:54.791838Z","iopub.status.idle":"2024-09-07T14:53:54.799084Z","shell.execute_reply.started":"2024-09-07T14:53:54.791811Z","shell.execute_reply":"2024-09-07T14:53:54.798088Z"},"trusted":true},"execution_count":103,"outputs":[]},{"cell_type":"markdown","source":"words","metadata":{}},{"cell_type":"code","source":"# Weights export for words Bigram\ntorch.save(mwb.state_dict(), 'model_words_weights_bigram.pth')","metadata":{"execution":{"iopub.status.busy":"2024-08-31T07:22:40.728066Z","iopub.execute_input":"2024-08-31T07:22:40.728911Z","iopub.status.idle":"2024-08-31T07:22:41.492504Z","shell.execute_reply.started":"2024-08-31T07:22:40.728879Z","shell.execute_reply":"2024-08-31T07:22:41.491362Z"},"trusted":true},"execution_count":163,"outputs":[]},{"cell_type":"markdown","source":"## GPT","metadata":{}},{"cell_type":"markdown","source":"chars","metadata":{}},{"cell_type":"code","source":"# Weights export for chars GPT\ntorch.save(mc.state_dict(), 'model_chars_weights.pth')","metadata":{"execution":{"iopub.status.busy":"2024-08-27T14:19:13.412627Z","iopub.execute_input":"2024-08-27T14:19:13.413401Z","iopub.status.idle":"2024-08-27T14:19:14.100362Z","shell.execute_reply.started":"2024-08-27T14:19:13.413367Z","shell.execute_reply":"2024-08-27T14:19:14.099364Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"words","metadata":{}},{"cell_type":"code","source":"# Weights export for words GPT\ntorch.save(mw.state_dict(), 'model_words_weights.pth')","metadata":{"execution":{"iopub.status.busy":"2024-09-11T18:47:04.581210Z","iopub.execute_input":"2024-09-11T18:47:04.581704Z","iopub.status.idle":"2024-09-11T18:47:05.494275Z","shell.execute_reply.started":"2024-09-11T18:47:04.581660Z","shell.execute_reply":"2024-09-11T18:47:05.493230Z"},"trusted":true},"execution_count":45,"outputs":[]},{"cell_type":"markdown","source":"## Weights loading for GPT","metadata":{}},{"cell_type":"markdown","source":"chars","metadata":{}},{"cell_type":"code","source":"# Weights loading for chars\nmc.load_state_dict(torch.load('/kaggle/input/transformer-weights/model_chars_weights.pth'))","metadata":{"execution":{"iopub.status.busy":"2024-09-26T13:18:09.856833Z","iopub.execute_input":"2024-09-26T13:18:09.857189Z","iopub.status.idle":"2024-09-26T13:18:11.956702Z","shell.execute_reply.started":"2024-09-26T13:18:09.857163Z","shell.execute_reply":"2024-09-26T13:18:11.955749Z"},"trusted":true},"execution_count":17,"outputs":[{"execution_count":17,"output_type":"execute_result","data":{"text/plain":"<All keys matched successfully>"},"metadata":{}}]},{"cell_type":"markdown","source":"words","metadata":{}},{"cell_type":"code","source":"# Weights loading for words\nmw.load_state_dict(torch.load('/kaggle/input/llm-gpt/pytorch/default/1/model_words_weights BEST.pth'))","metadata":{"execution":{"iopub.status.busy":"2024-09-30T13:07:04.714374Z","iopub.execute_input":"2024-09-30T13:07:04.715166Z","iopub.status.idle":"2024-09-30T13:07:07.332739Z","shell.execute_reply.started":"2024-09-30T13:07:04.715132Z","shell.execute_reply":"2024-09-30T13:07:07.331987Z"},"trusted":true},"execution_count":15,"outputs":[{"execution_count":15,"output_type":"execute_result","data":{"text/plain":"<All keys matched successfully>"},"metadata":{}}]}]}