QA-Shlyahin-submit-Parts-5-6.ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"colab":{"name":"QA-Shlyahin-submit-Parts-5-6.ipynb","provenance":[],"collapsed_sections":[]}},"nbformat_minor":0,"nbformat":4,"cells":[{"cell_type":"markdown","source":["# Проект QA\n","\n","## Yes/No Questions\n","\n","Вы будете работать с корпусом BoolQ. Корпус состоит из вопросов, предполагающих бинарный ответ (да / нет), абзацев из Википедии,  содержащих ответ на вопрос, заголовка статьи, из которой извлечен абзац и непосредственно ответа (true / false).\n","\n","Корпус описан в статье:\n","\n","Christopher Clark, Kenton Lee, Ming-Wei Chang, Tom Kwiatkowski, Michael Collins, Kristina Toutanova\n","BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions\n","\n","https://arxiv.org/abs/1905.10044\n","\n","\n","Корпус (train-dev split) доступен в репозитории проекта:  https://github.com/google-research-datasets/boolean-questions\n","\n","Используйте для обучения train часть корпуса, для валидации и тестирования – dev часть. \n","\n","Каждый бонус пункт оцениватся в 1 балл. Вывод являяется обязательным!"],"metadata":{"id":"Fd06vTCzqj3y"}},{"cell_type":"code","source":["!gsutil cp gs://boolq/train.jsonl .\n","!gsutil cp gs://boolq/dev.jsonl ./"],"metadata":{"id":"FOup-Ca-q5Ay","executionInfo":{"status":"ok","timestamp":1644744135541,"user_tz":-180,"elapsed":4431,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"c79ab92e-aec3-4d06-8f47-eb11e83fda1d","execution":{"iopub.status.busy":"2022-02-13T11:15:03.697885Z","iopub.execute_input":"2022-02-13T11:15:03.699029Z","iopub.status.idle":"2022-02-13T11:15:15.552729Z","shell.execute_reply.started":"2022-02-13T11:15:03.698867Z","shell.execute_reply":"2022-02-13T11:15:15.551475Z"},"trusted":true},"execution_count":null,"outputs":[{"name":"stdout","text":"Copying gs://boolq/train.jsonl...\n\\ [1 files][  6.2 MiB/  6.2 MiB]                                                \nOperation completed over 1 objects/6.2 MiB.                                      \nCopying gs://boolq/dev.jsonl...\n/ [1 files][  2.1 MiB/  2.1 MiB]                                                \nOperation completed over 1 objects/2.1 MiB.                                      \n","output_type":"stream"}]},{"cell_type":"code","source":["train_path = \"./train.jsonl\"\n","dev_path = \"./dev.jsonl\""],"metadata":{"id":"k5hixjuOzM3w","execution":{"iopub.status.busy":"2022-02-13T11:15:15.555652Z","iopub.execute_input":"2022-02-13T11:15:15.556226Z","iopub.status.idle":"2022-02-13T11:15:15.564090Z","shell.execute_reply.started":"2022-02-13T11:15:15.556179Z","shell.execute_reply":"2022-02-13T11:15:15.563222Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["!pip install transformers"],"metadata":{"id":"AkgNpO5vkXVd","executionInfo":{"status":"ok","timestamp":1644744139043,"user_tz":-180,"elapsed":3506,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"8d3a035d-9a8a-4faa-f221-ce3e76268e51","execution":{"iopub.status.busy":"2022-02-13T11:15:15.567578Z","iopub.execute_input":"2022-02-13T11:15:15.568623Z","iopub.status.idle":"2022-02-13T11:15:25.428059Z","shell.execute_reply.started":"2022-02-13T11:15:15.568574Z","shell.execute_reply":"2022-02-13T11:15:25.426983Z"},"trusted":true},"execution_count":null,"outputs":[{"name":"stdout","text":"Requirement already satisfied: transformers in /opt/conda/lib/python3.7/site-packages (4.15.0)\nRequirement already satisfied: tqdm>=4.27 in /opt/conda/lib/python3.7/site-packages (from transformers) (4.62.3)\nRequirement already satisfied: pyyaml>=5.1 in /opt/conda/lib/python3.7/site-packages (from transformers) (6.0)\nRequirement already satisfied: requests in /opt/conda/lib/python3.7/site-packages (from transformers) (2.26.0)\nRequirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from transformers) (4.10.1)\nRequirement already satisfied: packaging>=20.0 in /opt/conda/lib/python3.7/site-packages (from transformers) (21.3)\nRequirement already satisfied: regex!=2019.12.17 in /opt/conda/lib/python3.7/site-packages (from transformers) (2021.11.10)\nRequirement already satisfied: sacremoses in /opt/conda/lib/python3.7/site-packages (from transformers) (0.0.47)\nRequirement already satisfied: tokenizers<0.11,>=0.10.1 in /opt/conda/lib/python3.7/site-packages (from transformers) (0.10.3)\nRequirement already satisfied: filelock in /opt/conda/lib/python3.7/site-packages (from transformers) (3.4.2)\nRequirement already satisfied: numpy>=1.17 in /opt/conda/lib/python3.7/site-packages (from transformers) (1.20.3)\nRequirement already satisfied: huggingface-hub<1.0,>=0.1.0 in /opt/conda/lib/python3.7/site-packages (from transformers) (0.2.1)\nRequirement already satisfied: typing-extensions>=3.7.4.3 in /opt/conda/lib/python3.7/site-packages (from huggingface-hub<1.0,>=0.1.0->transformers) (4.0.1)\nRequirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /opt/conda/lib/python3.7/site-packages (from packaging>=20.0->transformers) (3.0.6)\nRequirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->transformers) (3.6.0)\nRequirement already satisfied: charset-normalizer~=2.0.0 in /opt/conda/lib/python3.7/site-packages (from requests->transformers) (2.0.9)\nRequirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests->transformers) (2021.10.8)\nRequirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests->transformers) (3.1)\nRequirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests->transformers) (1.26.7)\nRequirement already satisfied: six in /opt/conda/lib/python3.7/site-packages (from sacremoses->transformers) (1.16.0)\nRequirement already satisfied: joblib in /opt/conda/lib/python3.7/site-packages (from sacremoses->transformers) (1.1.0)\nRequirement already satisfied: click in /opt/conda/lib/python3.7/site-packages (from sacremoses->transformers) (8.0.3)\n^C\n\u001b[31mERROR: Operation cancelled by user\u001b[0m\n","output_type":"stream"}]},{"cell_type":"code","source":["import random\n","import torch\n","from torch import nn\n","import numpy as np\n","import pandas as pd\n","from tqdm import tqdm\n","from torch.utils.data import TensorDataset, RandomSampler, SequentialSampler\n","# from transformers import AutoTokenizer, AutoModelForSequenceClassification, AdamW\n","from matplotlib import pyplot as plt\n","from sklearn.metrics import accuracy_score, f1_score\n","import warnings\n","warnings.filterwarnings('ignore')\n","\n","from sklearn.linear_model import LogisticRegression\n","from sklearn.model_selection import GridSearchCV\n","from sklearn.model_selection import cross_val_score\n","import transformers as ppb\n","\n","import re\n","\n","from sklearn.model_selection import train_test_split\n","\n","from tqdm.notebook import tqdm\n","\n","from torch.utils.data import Dataset, random_split, Subset, DataLoader, Sampler\n","\n","import torch.optim as optim\n","\n","import time\n","import math\n","import matplotlib\n","matplotlib.rcParams.update({'figure.figsize': (16, 12), 'font.size': 14})\n","%matplotlib inline\n","from IPython.display import clear_output\n","\n","import gc\n","\n","import json\n","import os\n","import nltk\n","import torch\n","\n","from torchtext.legacy import data\n","from torchtext.legacy import datasets\n","from torchtext.vocab import GloVe\n","\n","\n","from nltk.tokenize.stanford import StanfordTokenizer\n","\n","import torch.nn as nn\n","import torch.nn.functional as F\n","\n","\n","from dataclasses import dataclass, field"],"metadata":{"id":"_fizTGuBsaPr","execution":{"iopub.status.busy":"2022-02-13T11:15:25.431285Z","iopub.execute_input":"2022-02-13T11:15:25.431772Z","iopub.status.idle":"2022-02-13T11:15:28.973256Z","shell.execute_reply.started":"2022-02-13T11:15:25.431726Z","shell.execute_reply":"2022-02-13T11:15:28.972211Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n","\n","# зафиксируем рандомы для воспроизводимости результатов\n","\n","random.seed(26)\n","np.random.seed(26)\n","torch.manual_seed(26)"],"metadata":{"id":"x4TUDFtUypbn","executionInfo":{"status":"ok","timestamp":1644744139044,"user_tz":-180,"elapsed":9,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"6b9c109b-93be-4f7f-dc67-99d398df93a4","execution":{"iopub.status.busy":"2022-02-13T11:15:28.975244Z","iopub.execute_input":"2022-02-13T11:15:28.975629Z","iopub.status.idle":"2022-02-13T11:15:29.041292Z","shell.execute_reply.started":"2022-02-13T11:15:28.975585Z","shell.execute_reply":"2022-02-13T11:15:29.040093Z"},"trusted":true},"execution_count":null,"outputs":[{"execution_count":5,"output_type":"execute_result","data":{"text/plain":"<torch._C.Generator at 0x7fbd356e0490>"},"metadata":{}}]},{"cell_type":"code","source":["train_data_df = pd.read_json(\"train.jsonl\", lines=True, orient='records')\n","dev_data_df = pd.read_json(\"dev.jsonl\", lines=True, orient=\"records\")\n","\n","passages_train = train_data_df.passage.values\n","questions_train = train_data_df.question.values\n","answers_train = train_data_df.answer.values.astype(int)\n","\n","passages_dev = dev_data_df.passage.values\n","questions_dev = dev_data_df.question.values\n","answers_dev = dev_data_df.answer.values.astype(int)"],"metadata":{"id":"gZ1V6E5xyv71","execution":{"iopub.status.busy":"2022-02-13T11:15:29.043122Z","iopub.execute_input":"2022-02-13T11:15:29.043512Z","iopub.status.idle":"2022-02-13T11:15:29.219125Z","shell.execute_reply.started":"2022-02-13T11:15:29.043443Z","shell.execute_reply":"2022-02-13T11:15:29.218120Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["questions_dev[-5:]"],"metadata":{"id":"3PHT3C2Kzm_H","executionInfo":{"status":"ok","timestamp":1644744139431,"user_tz":-180,"elapsed":5,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"a88a03ac-a4bc-4e42-bee9-4ef1bc2a7578","execution":{"iopub.status.busy":"2022-02-13T11:15:29.220898Z","iopub.execute_input":"2022-02-13T11:15:29.221237Z","iopub.status.idle":"2022-02-13T11:15:29.229601Z","shell.execute_reply.started":"2022-02-13T11:15:29.221192Z","shell.execute_reply":"2022-02-13T11:15:29.228406Z"},"trusted":true},"execution_count":null,"outputs":[{"execution_count":7,"output_type":"execute_result","data":{"text/plain":"array(['is manic depression the same as bi polar',\n       'was whiskey galore based on a true story',\n       'are there plants on the international space station',\n       'does the hockey puck have to cross the line to be a goal',\n       'will there be a season 5 of shadowhunters'], dtype=object)"},"metadata":{}}]},{"cell_type":"markdown","source":["Для проверки моделей приготовим тестовыe примеры."],"metadata":{"id":"jx6BjKV5JjzM"}},{"cell_type":"code","source":["test_passages = passages_dev[-5:]\n","test_questions = questions_dev[-5:]\n","test_answers = answers_dev[-5:]"],"metadata":{"id":"WWI6XZ96zWV5","execution":{"iopub.status.busy":"2022-02-13T11:15:29.231670Z","iopub.execute_input":"2022-02-13T11:15:29.232477Z","iopub.status.idle":"2022-02-13T11:15:29.240759Z","shell.execute_reply.started":"2022-02-13T11:15:29.232414Z","shell.execute_reply":"2022-02-13T11:15:29.239541Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Пример вопроса: \n","question: is batman and robin a sequel to batman forever\n","\n","title: Batman & Robin (film)\n","\n","answer: true\n","\n","passage: With the box office success of Batman Forever in June 1995, Warner Bros. immediately commissioned a sequel. They hired director Joel Schumacher and writer Akiva Goldsman to reprise their duties the following August, and decided it was best to fast track production for a June 1997 target release date, which is a break from the usual 3-year gap between films. Schumacher wanted to homage both the broad camp style of the 1960s television series and the work of Dick Sprang. The storyline of Batman & Robin was conceived by Schumacher and Goldsman during pre-production on A Time to Kill. Portions of Mr. Freeze's back-story were based on the Batman: The Animated Series episode ''Heart of Ice'', written by Paul Dini."],"metadata":{"id":"yDEsCCkoqj34"}},{"cell_type":"markdown","source":["## Часть 5. BiDAF-подобная архитектура (2 балла)\n","\n","Архитектура BiDAF является специфической архитектурой для QA. Здесь вам предлагается её реализовать.\n","\n","Основана на статье: Bidirectional Attention Flow for Machine Comprehension\n","\n","Minjoon Seo, Aniruddha Kembhavi, Ali Farhadi, Hannaneh Hajishirzi\n","\n","https://arxiv.org/abs/1611.01603\n","\n","Архитектура BiDAF предложена для задачи SQuAD, но легко может быть адаптирована к текущему заданию. Модель состоит из следующих блоков:\n","1. Кодировщик  получает на вход два представления слова: эмбеддинг слова и полученное из CNN посимвольное представление слова. Кодировщики для вопроса и для параграфа одинаковы. \n","2. Слой внимания (детальное описание приведено в статье, см. пункт Attention Flow Layer)\n","3. Промежуточный слой, который получает на вход контекстуализированные эмбеддинги слов из параграфа, состоящие из трех частей (выход кодировщика параграфа,   Query2Context (один вектор) и Context2Query (матрица) выравнивания\n","\n","4. Слой предсказания. \n","\n","Предложите, как можно было модифицировать последний слой предсказания в архитектуре BiDAF, с учетом того, что итоговое предсказание – это метка yes / no, предсказание которой проще, чем предсказание спана ответа для SQuAD.\n","\n","Оцените качество этой модели для решения задачи. \n","\n","[bonus] Замените входные эмбеддинги и все дополнительные признаки, используемые кодировщиками, на BERT или RoBERTa эмбеддинги. Улучшит ли это качество результатов?"],"metadata":{"id":"MXmluyTFqj4P"}},{"cell_type":"markdown","source":["Сравнение DrQA и BiDAF:\n","    \n","![](https://www.researchgate.net/profile/Felix_Wu6/publication/321069852/figure/fig1/AS:560800147881984@1510716582560/Schematic-layouts-of-the-BiDAF-left-and-DrQA-right-architectures-We-propose-to.png)"],"metadata":{"id":"FrPAxeHIqj4S"}},{"cell_type":"markdown","source":["Для проекта я адаптировал код BiDAF из репозитория https://github.com/galsang/BiDAF-pytorch"],"metadata":{"id":"G9IffQR9zM4X"}},{"cell_type":"markdown","source":["Подготовим данные и итераторы, используя torchtext. С описанным в статье PTBTokenizer'ом результаты были хуже. Поэтому возьмем word_tokenizer из nltk, как сделано в вышеуказанном репозитории."],"metadata":{"id":"2j4NVe5mN6Yt"}},{"cell_type":"code","source":[" nltk.download('punkt')"],"metadata":{"id":"hkboCiRTzM4X","executionInfo":{"status":"ok","timestamp":1644740121554,"user_tz":-180,"elapsed":699,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"df83c637-8129-41eb-c740-269c1461d582","execution":{"iopub.status.busy":"2022-02-13T11:15:29.243025Z","iopub.execute_input":"2022-02-13T11:15:29.243463Z","iopub.status.idle":"2022-02-13T11:15:29.458281Z","shell.execute_reply.started":"2022-02-13T11:15:29.243415Z","shell.execute_reply":"2022-02-13T11:15:29.457068Z"},"trusted":true},"execution_count":null,"outputs":[{"name":"stdout","text":"[nltk_data] Downloading package punkt to /usr/share/nltk_data...\n[nltk_data]   Package punkt is already up-to-date!\n","output_type":"stream"},{"execution_count":9,"output_type":"execute_result","data":{"text/plain":"True"},"metadata":{}}]},{"cell_type":"code","source":["def word_tokenize(tokens):\n","    return [token.replace(\"''\", '\"').replace(\"``\", '\"') for token in nltk.word_tokenize(tokens)]"],"metadata":{"id":"785AlUHpzM4X","execution":{"iopub.status.busy":"2022-02-13T11:15:29.463065Z","iopub.execute_input":"2022-02-13T11:15:29.463413Z","iopub.status.idle":"2022-02-13T11:15:29.469456Z","shell.execute_reply.started":"2022-02-13T11:15:29.463366Z","shell.execute_reply":"2022-02-13T11:15:29.468377Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["CHAR_NESTING = data.Field(batch_first=True, tokenize=list, lower=True)\n","CHAR = data.NestedField(CHAR_NESTING, tokenize=word_tokenize)\n","WORD = data.Field(batch_first=True, tokenize=word_tokenize, lower=True, include_lengths=True)\n","LABEL = data.Field(sequential=False, unk_token=None, use_vocab=False)\n","\n","dict_fields = {'answer': ('answer', LABEL),\n","                'passage': [('c_word', WORD), ('c_char', CHAR)],\n","                'question': [('q_word', WORD), ('q_char', CHAR)]}\n","\n","list_fields = [('answer', LABEL), \n","                ('c_word', WORD), ('c_char', CHAR),\n","                ('q_word', WORD), ('q_char', CHAR)]\n","\n","train, dev = data.TabularDataset.splits(path=\"\",  train=\"./train.jsonl\",\n","    validation=\"./dev.jsonl\",  format='json',  fields=dict_fields)"],"metadata":{"id":"WQX52rZxzM4X","execution":{"iopub.status.busy":"2022-02-13T11:15:57.775950Z","iopub.execute_input":"2022-02-13T11:15:57.776266Z","iopub.status.idle":"2022-02-13T11:16:46.853385Z","shell.execute_reply.started":"2022-02-13T11:15:57.776235Z","shell.execute_reply":"2022-02-13T11:16:46.852229Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["train = data.Dataset(examples=train, fields=list_fields)\n","dev = data.Dataset(examples=dev, fields=list_fields)\n","val, test = dev.split(split_ratio=0.5)"],"metadata":{"id":"Zs3gA6ggzM4Y","execution":{"iopub.status.busy":"2022-02-13T11:16:46.855748Z","iopub.execute_input":"2022-02-13T11:16:46.856067Z","iopub.status.idle":"2022-02-13T11:16:46.869736Z","shell.execute_reply.started":"2022-02-13T11:16:46.856007Z","shell.execute_reply":"2022-02-13T11:16:46.866918Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["d = 100\n","batch_size = 8\n","CHAR.build_vocab(train, val, test)\n","WORD.build_vocab(train, val, test, vectors=GloVe(name='6B', dim=d))\n","\n","train_iter = data.BucketIterator(\n","    train,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    shuffle=True,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")\n","\n","val_iter = data.BucketIterator(\n","    val,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")\n","\n","test_iter = data.BucketIterator(\n","    test,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")"],"metadata":{"id":"GnkI5U77zM4Y","executionInfo":{"status":"ok","timestamp":1644740349100,"user_tz":-180,"elapsed":199526,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"c306a407-2437-48bb-cb25-c8ddc6025faa","execution":{"iopub.status.busy":"2022-02-13T11:16:46.871643Z","iopub.execute_input":"2022-02-13T11:16:46.872129Z","iopub.status.idle":"2022-02-13T11:20:25.038185Z","shell.execute_reply.started":"2022-02-13T11:16:46.872077Z","shell.execute_reply":"2022-02-13T11:20:25.036762Z"},"trusted":true},"execution_count":null,"outputs":[{"name":"stderr","text":".vector_cache/glove.6B.zip: 862MB [02:43, 5.29MB/s]                               \n100%|█████████▉| 399999/400000 [00:26<00:00, 14845.17it/s]\n","output_type":"stream"}]},{"cell_type":"code","source":["char_vocab_size = len(CHAR.vocab)\n","word_vocab_size = len(WORD.vocab)"],"metadata":{"id":"iqhG0ixgzM4Y","execution":{"iopub.status.busy":"2022-02-13T11:20:25.041201Z","iopub.execute_input":"2022-02-13T11:20:25.041849Z","iopub.status.idle":"2022-02-13T11:20:25.048237Z","shell.execute_reply.started":"2022-02-13T11:20:25.041785Z","shell.execute_reply":"2022-02-13T11:20:25.046846Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Определим функции обучения и тестирования."],"metadata":{"id":"aWoCtXagOdiR"}},{"cell_type":"code","source":["def train(model, iterator, optimizer, criterion, scheduler, clip, train_history=None, valid_history=None):\n","    model.train()\n","    \n","    epoch_loss = 0\n","    history = []\n","    \n","    for i, batch in enumerate(iterator):\n","\n","        inputs = batch #.to(device)\n","        labels = batch.answer.float().to(device)\n","        optimizer.zero_grad()\n","\n","        output = model(inputs).squeeze(1)\n","\n","        loss = criterion(output, labels) \n","        loss.backward()\n","        torch.nn.utils.clip_grad_norm_(model.parameters(), clip)\n","        optimizer.step()\n","        if scheduler is not None:\n","            scheduler.step()\n","\n","        epoch_loss += loss.item()\n","        \n","        history.append(loss.cpu().data.numpy())\n","        \n","        if (i+1)%10==0:\n","            fig, ax = plt.subplots(nrows=1, ncols=2, figsize=(12, 8))\n","\n","            clear_output(True)\n","            ax[0].plot(history, label='train loss')\n","            ax[0].set_xlabel('Batch')\n","            ax[0].set_title('Train loss')\n","            if train_history is not None:\n","                ax[1].plot(train_history, label='general train history')\n","                ax[1].set_xlabel('Epoch')\n","            if valid_history is not None:\n","                ax[1].plot(valid_history, label='general valid history')\n","            plt.legend()\n","            \n","            plt.show()\n","    return epoch_loss / (i + 1)\n","\n","def evaluate(model, iterator, criterion):\n","    \n","    model.eval()\n","    \n","    epoch_loss = 0\n","    \n","    history = []\n","    \n","    with torch.no_grad():\n","        \n","    \n","        for i, batch in enumerate(iterator):\n","            inputs = batch #.to(device)\n","            labels = batch.answer.float().to(device)\n","            output = model(inputs).squeeze(1)\n","            loss = criterion(output, labels)\n","            epoch_loss += loss.item()\n","        \n","    return epoch_loss / (i + 1)\n","\n","def epoch_time(start_time, end_time):\n","    elapsed_time = end_time - start_time\n","    elapsed_mins = int(elapsed_time / 60)\n","    elapsed_secs = int(elapsed_time - (elapsed_mins * 60))\n","    return elapsed_mins, elapsed_secs\n","\n","\n","def test (model, weights='best-val-model.pt', iterator=test_iter):\n","    model.load_state_dict(torch.load(weights))\n","    model.eval()\n","    test_acc = 0\n","    test_f1 = 0\n","    with torch.no_grad():\n","        for i, batch in enumerate(iterator):\n","            inputs = batch\n","            labels = batch.answer.float().cpu().numpy()\n","            output = model(inputs).squeeze(1).cpu().detach().numpy() > 0.5\n","            test_acc += accuracy_score(labels, output)\n","            test_f1 += f1_score(labels, output)\n","    return (test_acc / (i + 1), test_f1 / (i + 1))"],"metadata":{"id":"S77Q5SCPzM4Y","execution":{"iopub.status.busy":"2022-02-13T11:20:25.051966Z","iopub.execute_input":"2022-02-13T11:20:25.053397Z","iopub.status.idle":"2022-02-13T11:20:25.091429Z","shell.execute_reply.started":"2022-02-13T11:20:25.053347Z","shell.execute_reply":"2022-02-13T11:20:25.090147Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["def train_epochs (clf, train_loader, valid_loader, optimizer, criterion, scheduler, CLIP=1, N_EPOCHS=3, patience=3):\n","    \n","    train_history = []\n","    valid_history = []\n","    \n","    best_valid_loss = float('inf')\n","\n","    for epoch in range(N_EPOCHS):\n","        \n","        torch.cuda.empty_cache()\n","        gc.collect()\n","\n","        start_time = time.time()\n","        \n","        train_loss = train(clf, train_loader, optimizer, criterion, scheduler, CLIP, train_history, valid_history)\n","        \n","        torch.cuda.empty_cache()\n","        gc.collect()\n","        \n","        valid_loss = evaluate(clf, valid_loader, criterion)\n","        \n","        end_time = time.time()\n","        epoch_mins, epoch_secs = epoch_time(start_time, end_time)\n","        \n","        if best_valid_loss > valid_loss:\n","            best_valid_loss = valid_loss\n","            torch.save(model.state_dict(), 'best-val-model.pt')\n","            cnt = 0\n","        else:\n","            cnt += 1\n","\n","        if cnt > patience:\n","            break\n","\n","\n","        train_history.append(train_loss)\n","        valid_history.append(valid_loss)\n","        print(f'Epoch: {epoch+1:02} | Time: {epoch_mins}m {epoch_secs}s')\n","        print(f'\\tTrain Loss: {train_loss:.3f} | Train PPL: {math.exp(train_loss):7.3f}')\n","        print(f'\\t Val. Loss: {valid_loss:.3f} |  Val. PPL: {math.exp(valid_loss):7.3f}')\n","        \n","    return train_history, valid_history"],"metadata":{"id":"A6pwviUOzM4Y","execution":{"iopub.status.busy":"2022-02-13T11:20:25.093760Z","iopub.execute_input":"2022-02-13T11:20:25.094612Z","iopub.status.idle":"2022-02-13T11:20:25.112321Z","shell.execute_reply.started":"2022-02-13T11:20:25.094547Z","shell.execute_reply":"2022-02-13T11:20:25.110946Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Адаптируем модель - возьмем скрытые состояния последнего слоя lstm из  Output layer (на эту lstm подаем конкатенацию эттеншенов и выхода Modeling layer). Эти скрытые состояния пропустим через один полносвязный слой-классификатор."],"metadata":{"id":"rD7N1kZbOjhM"}},{"cell_type":"code","source":["@dataclass\n","class Arguments():\n","  char_dim: int = 8\n","  char_channel_width: int = 5\n","  char_channel_size: int = 100 # 768 \n","  dropout: float = 0.2\n","  hidden_size: int = 100 #768 \n","  word_dim: int = 100 #768\n","  char_vocab_size: int = len(CHAR.vocab)\n","  word_vocab_size: int = len(WORD.vocab)\n","\n","args = Arguments()"],"metadata":{"id":"1p4zjS-RzM4Z","execution":{"iopub.status.busy":"2022-02-13T11:20:26.049772Z","iopub.execute_input":"2022-02-13T11:20:26.050265Z","iopub.status.idle":"2022-02-13T11:20:26.060877Z","shell.execute_reply.started":"2022-02-13T11:20:26.050216Z","shell.execute_reply":"2022-02-13T11:20:26.059739Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["class LSTM(nn.Module):\n","    def __init__(self, input_size, hidden_size, batch_first=False, num_layers=1, bidirectional=False, dropout=0.2):\n","        super(LSTM, self).__init__()\n","\n","        self.rnn = nn.LSTM(input_size=input_size,\n","                           hidden_size=hidden_size,\n","                           num_layers=num_layers,\n","                           bidirectional=bidirectional,\n","                           batch_first=batch_first)\n","\n","        self.dropout = nn.Dropout(p=dropout)\n","\n","\n","\n","    def forward(self, x):\n","        x, x_len = x\n","        x = self.dropout(x)\n","\n","        x_len_sorted, x_idx = torch.sort(x_len, descending=True)\n","        x_sorted = x.index_select(dim=0, index=x_idx)\n","        _, x_ori_idx = torch.sort(x_idx)\n","\n","        x_packed = nn.utils.rnn.pack_padded_sequence(x_sorted, x_len_sorted.to('cpu'), batch_first=True)\n","        x_packed, (h, c) = self.rnn(x_packed)\n","\n","        x = nn.utils.rnn.pad_packed_sequence(x_packed, batch_first=True)[0]\n","        x = x.index_select(dim=0, index=x_ori_idx)\n","        h = h.permute(1, 0, 2).contiguous().view(-1, h.size(0) * h.size(2)).squeeze()\n","        h = h.index_select(dim=0, index=x_ori_idx)\n","\n","        return x, h\n","\n","\n","class Linear(nn.Module):\n","    def __init__(self, in_features, out_features, dropout=0.0):\n","        super(Linear, self).__init__()\n","\n","        self.linear = nn.Linear(in_features=in_features, out_features=out_features)\n","        if dropout > 0:\n","            self.dropout = nn.Dropout(p=dropout)\n","        self.reset_params()\n","\n","    def reset_params(self):\n","        nn.init.kaiming_normal_(self.linear.weight)\n","        nn.init.constant_(self.linear.bias, 0)\n","\n","    def forward(self, x):\n","        if hasattr(self, 'dropout'):\n","            x = self.dropout(x)\n","        x = self.linear(x)\n","        return x\n","\n","class BiDAF2(nn.Module):\n","    def __init__(self, args, pretrained):\n","        super(BiDAF2, self).__init__()\n","        self.args = args\n","\n","        # 1. Character Embedding Layer\n","        self.char_emb = nn.Embedding(args.char_vocab_size, args.char_dim, padding_idx=1)\n","        # self.char_emb = nn.Embedding.from_pretrained(pretrained, freeze=True)\n","        nn.init.uniform_(self.char_emb.weight, -0.001, 0.001)\n","\n","        self.char_conv = nn.Sequential(\n","            nn.Conv2d(1, args.char_channel_size, (args.char_dim, args.char_channel_width)),\n","            nn.ReLU()\n","            )\n","\n","        # 2. Word Embedding Layer\n","        # initialize word embedding with GloVe\n","        self.word_emb = nn.Embedding.from_pretrained(pretrained, freeze=True)\n","\n","        # highway network\n","        assert self.args.hidden_size * 2 == (self.args.char_channel_size + self.args.word_dim)\n","        for i in range(2):\n","            setattr(self, 'highway_linear{}'.format(i),\n","                    nn.Sequential(Linear(args.hidden_size * 2, args.hidden_size * 2),\n","                                  nn.ReLU()))\n","            setattr(self, 'highway_gate{}'.format(i),\n","                    nn.Sequential(Linear(args.hidden_size * 2, args.hidden_size * 2),\n","                                  nn.Sigmoid()))\n","\n","        # 3. Contextual Embedding Layer\n","        self.context_LSTM = LSTM(input_size=args.hidden_size * 2,\n","                                 hidden_size=args.hidden_size,\n","                                 bidirectional=True,\n","                                 batch_first=True,\n","                                 dropout=args.dropout)\n","\n","        # 4. Attention Flow Layer\n","        self.att_weight_c = Linear(args.hidden_size * 2, 1)\n","        self.att_weight_q = Linear(args.hidden_size * 2, 1)\n","        self.att_weight_cq = Linear(args.hidden_size * 2, 1)\n","\n","        # 5. Modeling Layer\n","        self.LSTM1 = LSTM(input_size=args.hidden_size * 8, hidden_size=args.hidden_size,\n","                      bidirectional=True, batch_first=True, dropout=args.dropout)\n","        self.LSTM2 = LSTM(input_size=args.hidden_size * 2, hidden_size=args.hidden_size,\n","                      bidirectional=True, batch_first=True, dropout=args.dropout)\n","    \n","        # 6. Output Layer\n","        self.preclassifier = Linear(args.hidden_size * 2, args.hidden_size * 2, dropout=args.dropout)\n","        self.classifier = Linear(args.hidden_size * 2, 1, dropout=args.dropout)\n","        self.drop=nn.Dropout(p=args.dropout)\n","        self.sigmoid = nn.Sigmoid()\n","\n","        self.lstm = LSTM(\n","            input_size=10 * args.hidden_size,\n","            hidden_size=args.hidden_size,\n","            bidirectional=True,\n","            batch_first=True,\n","            dropout=args.dropout\n","        )\n","\n","        self.out = nn.Sequential(\n","            nn.Dropout(p=args.dropout),\n","            nn.Linear(2 * args.hidden_size, 1)\n","        )\n","\n","        self.dropout = nn.Dropout(p=args.dropout)\n","        \n","\n","    def forward(self, batch):\n","        \n","        def char_emb_layer(x):\n","            \"\"\"\n","            :param x: (batch, seq_len, word_len)\n","            :return: (batch, seq_len, char_channel_size)\n","            \"\"\"\n","            batch_size = x.size(0)\n","            # (batch, seq_len, word_len, char_dim)\n","            x = self.dropout(self.char_emb(x))\n","            # (batch， seq_len, char_dim, word_len)\n","            x = x.transpose(2, 3)\n","            # (batch * seq_len, 1, char_dim, word_len)\n","            x = x.view(-1, self.args.char_dim, x.size(3)).unsqueeze(1)\n","            # (batch * seq_len, char_channel_size, 1, conv_len) -> (batch * seq_len, char_channel_size, conv_len)\n","            x = self.char_conv(x).squeeze()\n","            # (batch * seq_len, char_channel_size, 1) -> (batch * seq_len, char_channel_size)\n","            x = F.max_pool1d(x, x.size(2)).squeeze()\n","            # (batch, seq_len, char_channel_size)\n","            x = x.view(batch_size, -1, self.args.char_channel_size)\n","\n","            return x\n","\n","        def highway_network(x1, x2):\n","            \"\"\"\n","            :param x1: (batch, seq_len, char_channel_size)\n","            :param x2: (batch, seq_len, word_dim)\n","            :return: (batch, seq_len, hidden_size * 2)\n","            \"\"\"\n","            # (batch, seq_len, char_channel_size + word_dim)\n","            x = torch.cat([x1, x2], dim=-1)\n","            for i in range(2):\n","                h = getattr(self, 'highway_linear{}'.format(i))(x)\n","                g = getattr(self, 'highway_gate{}'.format(i))(x)\n","                x = g * h + (1 - g) * x\n","            # (batch, seq_len, hidden_size * 2)\n","            return x\n","\n","        def att_flow_layer(c, q):\n","            \"\"\"\n","            :param c: (batch, c_len, hidden_size * 2)\n","            :param q: (batch, q_len, hidden_size * 2)\n","            :return: (batch, c_len, q_len)\n","            \"\"\"\n","            c_len = c.size(1)\n","            q_len = q.size(1)\n","\n","            cq = []\n","            for i in range(q_len):\n","                #(batch, 1, hidden_size * 2)\n","                qi = q.select(1, i).unsqueeze(1)\n","                #(batch, c_len, 1)\n","                ci = self.att_weight_cq(c * qi).squeeze()\n","                cq.append(ci)\n","            # (batch, c_len, q_len)\n","            cq = torch.stack(cq, dim=-1)\n","\n","            # (batch, c_len, q_len)\n","            s = self.att_weight_c(c).expand(-1, -1, q_len) + \\\n","                self.att_weight_q(q).permute(0, 2, 1).expand(-1, c_len, -1) + \\\n","                cq\n","\n","            # (batch, c_len, q_len)\n","            a = F.softmax(s, dim=2)\n","            # (batch, c_len, q_len) * (batch, q_len, hidden_size * 2) -> (batch, c_len, hidden_size * 2)\n","            c2q_att = torch.bmm(a, q)\n","            # (batch, 1, c_len)\n","            b = F.softmax(torch.max(s, dim=2)[0], dim=1).unsqueeze(1)\n","            # (batch, 1, c_len) * (batch, c_len, hidden_size * 2) -> (batch, hidden_size * 2)\n","            q2c_att = torch.bmm(b, c).squeeze()\n","            # (batch, c_len, hidden_size * 2) (tiled)\n","            q2c_att = q2c_att.unsqueeze(1).expand(-1, c_len, -1)\n","\n","\n","            # (batch, c_len, hidden_size * 8)\n","            x = torch.cat([c, c2q_att, c * c2q_att, c * q2c_att], dim=-1)\n","            return x\n","\n","\n","\n","        # 1. Character Embedding Layer\n","        c_char = char_emb_layer(batch.c_char.to(device))\n","        q_char = char_emb_layer(batch.q_char.to(device))\n","        # 2. Word Embedding Layer\n","        c_word = self.word_emb(batch.c_word[0].to(device))\n","        q_word = self.word_emb(batch.q_word[0].to(device))\n","        c_lens = batch.c_word[1]\n","        q_lens = batch.q_word[1]\n","\n","        # Highway network\n","        c = highway_network(c_char, c_word)\n","        q = highway_network(q_char, q_word)\n","\n","        # 3. Contextual Embedding Layer        \n","        c = self.context_LSTM((c, c_lens))[0]\n","        q = self.context_LSTM((q, q_lens))[0]\n","\n","        # 4. Attention Flow Layer\n","        g = att_flow_layer(c, q)\n","\n","        # 5. Modeling Layer\n","        out = self.LSTM1((g, c_lens))[0]\n","        m = self.LSTM2((out, c_lens))[0]\n","        \n","        # 6. Output Layer\n","        features = self.lstm((torch.cat([g, m], dim=-1), c_lens))\n","        p = self.sigmoid(self.out(features[1]))\n","    \n","        return p"],"metadata":{"id":"qy6KlcOPzM4Z","execution":{"iopub.status.busy":"2022-02-13T11:20:26.066276Z","iopub.execute_input":"2022-02-13T11:20:26.066579Z","iopub.status.idle":"2022-02-13T11:20:26.119613Z","shell.execute_reply.started":"2022-02-13T11:20:26.066544Z","shell.execute_reply":"2022-02-13T11:20:26.118535Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["torch.cuda.empty_cache()\n","gc.collect()\n","\n","n_epochs = 50\n","num_steps = len(train_iter) * n_epochs\n","args.dropout = 0.5\n","model = BiDAF2(args, pretrained=WORD.vocab.vectors).to(device)\n","\n","no_decay = ['bias', 'LayerNorm.weight']\n","decay = 0.01\n","optimizer_grouped_parameters = [\n","    {'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': decay},\n","    {'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}\n","]\n","optimizer = ppb.AdamW(optimizer_grouped_parameters, lr=5e-5, eps=1e-8)\n","\n","criterion = nn.BCELoss()\n","\n","train_epochs (model, train_iter, val_iter, optimizer, criterion, scheduler=None, \n","              CLIP=1, N_EPOCHS=n_epochs, patience=3) \n","print ('Test acc: ', test(BiDAF2(args, pretrained=WORD.vocab.vectors).to(device), \n","                          weights='best-val-model.pt', iterator=test_iter))"],"metadata":{"id":"TdTethzSzM4a","executionInfo":{"status":"ok","timestamp":1644742985628,"user_tz":-180,"elapsed":2634634,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"47770b78-50cb-4957-b072-bbef66f33b8f","execution":{"iopub.status.busy":"2022-02-13T11:20:26.122144Z","iopub.execute_input":"2022-02-13T11:20:26.122987Z","iopub.status.idle":"2022-02-13T11:48:45.921449Z","shell.execute_reply.started":"2022-02-13T11:20:26.122930Z","shell.execute_reply":"2022-02-13T11:48:45.920226Z"},"trusted":true},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 864x576 with 2 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"name":"stdout","text":"Test acc:  (0.6567073170731708, 0.7359855050098961)\n","output_type":"stream"}]},{"cell_type":"markdown","source":["Результат - 0.6567"],"metadata":{"id":"FMhQXNreLe-N"}},{"cell_type":"markdown","source":["Проверим модель на наших примерах."],"metadata":{"id":"oaIK2PwszM4d"}},{"cell_type":"code","source":["model = BiDAF2(args, pretrained=WORD.vocab.vectors).to(device)\n","model.load_state_dict(torch.load('best-val-model.pt'))"],"metadata":{"id":"fJX-9ztIWT4B","executionInfo":{"status":"ok","timestamp":1644743978057,"user_tz":-180,"elapsed":347,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"c8a1002e-9e11-4f5f-ce31-92bc93d14f89","execution":{"iopub.status.busy":"2022-02-13T11:48:45.947298Z","iopub.execute_input":"2022-02-13T11:48:45.948005Z","iopub.status.idle":"2022-02-13T11:48:46.023672Z","shell.execute_reply.started":"2022-02-13T11:48:45.947953Z","shell.execute_reply":"2022-02-13T11:48:46.021670Z"},"trusted":true},"execution_count":null,"outputs":[{"execution_count":25,"output_type":"execute_result","data":{"text/plain":"<All keys matched successfully>"},"metadata":{}}]},{"cell_type":"code","source":["def predict (model, questions, passages, answers, weights='best-val-model.pt'):\n","    \n","    examples = []\n","    for i in range(len(questions)):\n","        x = {'answer': answers[i], 'passage': passages[i], 'question': questions[i]}\n","        example = data.Example.fromdict(x, dict_fields)\n","        examples.append(example)\n","    our = data.Dataset(examples=examples, fields=list_fields)\n","    our_iter = data.BucketIterator(\n","    our,\n","    batch_size=len(our),\n","    device=device,\n","    repeat=False,\n","    shuffle=False,\n","    sort_within_batch = False\n","    )\n","\n","    model.load_state_dict(torch.load(weights))\n","    model.eval()\n","\n","    with torch.no_grad():\n","        for i, batch in enumerate(our_iter):\n","            inputs = batch#.to(device)\n","            labels = batch.answer.float().cpu().numpy()\n","            output = model(inputs).squeeze(1).cpu().detach().numpy() > 0.5\n","\n","    \n","    for i in range(len(labels)):\n","        print(\"Q.: \", questions[i])\n","        print(\"P.: \", passages[i])\n","        if output[i]:\n","            print ('YES')\n","        else:\n","            print('NO')"],"metadata":{"id":"OdjRNOqvzM4b","execution":{"iopub.status.busy":"2022-02-13T11:48:46.147765Z","iopub.execute_input":"2022-02-13T11:48:46.148468Z","iopub.status.idle":"2022-02-13T11:48:46.163167Z","shell.execute_reply.started":"2022-02-13T11:48:46.148413Z","shell.execute_reply":"2022-02-13T11:48:46.162062Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["predict (model, test_questions, test_passages, test_answers, weights='best-val-model.pt')"],"metadata":{"id":"xdVwBd8QzM4d","executionInfo":{"status":"ok","timestamp":1644744005711,"user_tz":-180,"elapsed":3,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"d17af901-e259-4f8d-ca5c-3ca662d70a08","execution":{"iopub.status.busy":"2022-02-13T11:48:46.165255Z","iopub.execute_input":"2022-02-13T11:48:46.165761Z","iopub.status.idle":"2022-02-13T11:48:46.264712Z","shell.execute_reply.started":"2022-02-13T11:48:46.165713Z","shell.execute_reply":"2022-02-13T11:48:46.263556Z"},"trusted":true},"execution_count":null,"outputs":[{"name":"stdout","text":"Q.:  is manic depression the same as bi polar\nP.:  Bipolar disorder, previously known as manic depression, is a mental disorder that causes periods of depression and periods of abnormally elevated mood. The elevated mood is significant and is known as mania or hypomania, depending on its severity, or whether symptoms of psychosis are present. During mania, an individual behaves or feels abnormally energetic, happy, or irritable. Individuals often make poorly thought out decisions with little regard to the consequences. The need for sleep is usually reduced during manic phases. During periods of depression, there may be crying, a negative outlook on life, and poor eye contact with others. The risk of suicide among those with the illness is high at greater than 6 percent over 20 years, while self-harm occurs in 30--40 percent. Other mental health issues such as anxiety disorders and substance use disorder are commonly associated.\nYES\nQ.:  was whiskey galore based on a true story\nP.:  SS Politician was an 8000-ton cargo ship owned by T & J Harrison of Liverpool. It left Liverpool on 3 February 1941, bound for Kingston, Jamaica and New Orleans with a cargo including 28,000 cases of malt whisky. The ship sank off the north coast of Eriskay in the Outer Hebrides, off the west coast of Scotland, and much of the wreck's cargo was salvaged by the island's inhabitants. The story of the wreck and looting was the basis for the book and film Whisky Galore!.\nYES\nQ.:  are there plants on the international space station\nP.:  Plant research continued on the International Space Station. Biomass Production System was used on the ISS Expedition 4. The Vegetable Production System (Veggie) system was later used aboard ISS. Plants tested in Veggie before going into space included lettuce, Swiss chard, radishes, Chinese cabbage and peas. Red Romaine lettuce was grown in space on Expedition 40 which were harvested when mature, frozen and tested back on Earth. Expedition 44 members became the first American astronauts to eat plants grown in space on 10 August 2015, when their crop of Red Romaine was harvested. Since 2003 Russian cosmonauts have been eating half of their crop while the other half goes towards further research. In 2012, a sunflower bloomed aboard the ISS under the care of NASA astronaut Donald Pettit. In January 2016, US astronauts announced that a zinnia had blossomed aboard the ISS.\nYES\nQ.:  does the hockey puck have to cross the line to be a goal\nP.:  In ice hockey, a goal is scored when the puck entirely crosses the goal line between the two goal posts and below the goal crossbar. A goal awards one point to the team attacking the goal scored upon, regardless of which team the player who actually deflected the puck into the goal belongs to (see also own goal). Typically, a player on the team attempting to score shoots the puck with his/her stick towards the goal net opening, and a player on the opposing team called a goaltender tries to block the shot to prevent a goal from being scored against his/her team.\nNO\nQ.:  will there be a season 5 of shadowhunters\nP.:  In April 2017, it was announced that the series had been renewed for a third season of 20 episodes. The first half of ten episodes premiered on March 20, 2018. In June 2018, Freeform canceled the series after three seasons, but ordered two extra episodes to properly conclude the series' story; the second half of the third season is set to air in early 2019. As of May 15, 2018, 43 episodes of Shadowhunters have aired, concluding the first half of the third season.\nNO\n","output_type":"stream"}]},{"cell_type":"markdown","source":["Результат на конкретных примерах лучше, чем у предыдущих моделей - на 4 из 5 вопросов ответ правильный."],"metadata":{"id":"maNVCgKZP-hM"}},{"cell_type":"markdown","source":["Теперь попробуем использовать эмбеддинги из BERT."],"metadata":{"id":"SBwC3k0zzM4b"}},{"cell_type":"code","source":["args.char_channel_size = 768 \n","args.dropout = 0.2\n","args.hidden_size = 768 \n","args.word_dim = 768\n","args.char_vocab_size = len(CHAR.vocab)\n","args.word_vocab_size = len(WORD.vocab)"],"metadata":{"id":"PhTQ56XBzM4b","execution":{"iopub.status.busy":"2022-02-13T11:48:46.269371Z","iopub.execute_input":"2022-02-13T11:48:46.269745Z","iopub.status.idle":"2022-02-13T11:48:46.275739Z","shell.execute_reply.started":"2022-02-13T11:48:46.269705Z","shell.execute_reply":"2022-02-13T11:48:46.274672Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["model_class, tokenizer_class, pretrained_weights = (ppb.BertModel, ppb.BertTokenizer, 'bert-base-uncased')\n","tokenizer = tokenizer_class.from_pretrained(pretrained_weights)\n","bert = model_class.from_pretrained(pretrained_weights).to(device)"],"metadata":{"id":"dlsqNcObzM4b","executionInfo":{"status":"ok","timestamp":1644744208134,"user_tz":-180,"elapsed":1892,"user":{"displayName":"Victor Shlyakhin","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"10685837151006696829"}},"outputId":"6dc0f452-4247-45a6-e7cb-57f03f55e5c6","execution":{"iopub.status.busy":"2022-02-13T11:48:46.277695Z","iopub.execute_input":"2022-02-13T11:48:46.278042Z","iopub.status.idle":"2022-02-13T11:49:11.809619Z","shell.execute_reply.started":"2022-02-13T11:48:46.277983Z","shell.execute_reply":"2022-02-13T11:49:11.808549Z"},"trusted":true,"colab":{"referenced_widgets":["4f9cac9395c74be8b2f1b20a38c73327","8835bf12dca2460f921bcd2335f9ea51","0fed56100e3641fbbddcb2f17c9b1dc3","0e218b3610d1404e81ea569cc116c2d1","97ce054ec71542f3ab0dd1046e6bc739"]}},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"Downloading:   0%|          | 0.00/226k [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"4f9cac9395c74be8b2f1b20a38c73327"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Downloading:   0%|          | 0.00/28.0 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"8835bf12dca2460f921bcd2335f9ea51"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Downloading:   0%|          | 0.00/455k [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"0fed56100e3641fbbddcb2f17c9b1dc3"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Downloading:   0%|          | 0.00/570 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"0e218b3610d1404e81ea569cc116c2d1"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Downloading:   0%|          | 0.00/420M [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"97ce054ec71542f3ab0dd1046e6bc739"}},"metadata":{}},{"name":"stderr","text":"Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.bias', 'cls.predictions.bias', 'cls.seq_relationship.weight']\n- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n","output_type":"stream"}]},{"cell_type":"code","source":["def bert_tokenize(sen):\n","    return tokenizer.tokenize(sen) "],"metadata":{"id":"PfDoaxx1zM4c","execution":{"iopub.status.busy":"2022-02-13T11:49:11.811549Z","iopub.execute_input":"2022-02-13T11:49:11.811929Z","iopub.status.idle":"2022-02-13T11:49:11.823582Z","shell.execute_reply.started":"2022-02-13T11:49:11.811863Z","shell.execute_reply":"2022-02-13T11:49:11.820854Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["CHAR_NESTING = data.Field(batch_first=True, tokenize=list, unk_token=tokenizer.unk_token)\n","CHAR = data.NestedField(CHAR_NESTING, tokenize=bert_tokenize)\n","WORD = data.Field(batch_first=True, sequential=True, tokenize=bert_tokenize, \n","                  pad_token=tokenizer.pad_token, unk_token=tokenizer.unk_token,\n","                  include_lengths=True)\n","LABEL = data.Field(sequential=False, unk_token=None, use_vocab=False)\n","\n","dict_fields = {'answer': ('answer', LABEL),\n","                'passage': [('c_word', WORD), ('c_char', CHAR)],\n","                'question': [('q_word', WORD), ('q_char', CHAR)]}\n","\n","list_fields = [('answer', LABEL), \n","                ('c_word', WORD), ('c_char', CHAR),\n","                ('q_word', WORD), ('q_char', CHAR)]\n","\n","train_set, dev_set = data.TabularDataset.splits(path=\"./\",  train=\"train.jsonl\",\n","   validation=\"dev.jsonl\",  format='json',  fields=dict_fields)"],"metadata":{"id":"266mbHn1zM4c","execution":{"iopub.status.busy":"2022-02-13T11:49:11.831306Z","iopub.execute_input":"2022-02-13T11:49:11.834064Z","iopub.status.idle":"2022-02-13T11:51:04.461779Z","shell.execute_reply.started":"2022-02-13T11:49:11.834010Z","shell.execute_reply":"2022-02-13T11:51:04.460732Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["train_dataset = data.Dataset(examples=train_set, fields=list_fields)\n","dev_dataset = data.Dataset(examples=dev_set, fields=list_fields)\n","val_dataset, test_dataset = dev_dataset.split(split_ratio=0.5)"],"metadata":{"id":"EiN44fT3zM4c","execution":{"iopub.status.busy":"2022-02-13T11:51:04.463236Z","iopub.execute_input":"2022-02-13T11:51:04.463579Z","iopub.status.idle":"2022-02-13T11:51:04.475821Z","shell.execute_reply.started":"2022-02-13T11:51:04.463492Z","shell.execute_reply":"2022-02-13T11:51:04.474764Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["CHAR.build_vocab(train_dataset, val_dataset, test_dataset)\n","WORD.build_vocab(train_dataset, val_dataset, test_dataset)\n","\n","WORD.vocab.stoi = tokenizer.vocab\n","WORD.vocab.itos = list(tokenizer.vocab)"],"metadata":{"id":"ek1YDBNRzM4c","execution":{"iopub.status.busy":"2022-02-13T11:51:04.477817Z","iopub.execute_input":"2022-02-13T11:51:04.478508Z","iopub.status.idle":"2022-02-13T11:51:08.486127Z","shell.execute_reply.started":"2022-02-13T11:51:04.478446Z","shell.execute_reply":"2022-02-13T11:51:08.485071Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["batch_size = 8\n","\n","\n","train_iter = data.BucketIterator(\n","    train_dataset,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    shuffle=True,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")\n","\n","val_iter = data.BucketIterator(\n","    val_dataset,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")\n","\n","test_iter = data.BucketIterator(\n","    test_dataset,\n","    batch_size=batch_size,\n","    device=device,\n","    repeat=False,\n","    sort_within_batch = True,\n","    sort_key=lambda x: len(x.c_word)\n",")"],"metadata":{"id":"w0brsM7EzM4c","execution":{"iopub.status.busy":"2022-02-13T11:51:08.487971Z","iopub.execute_input":"2022-02-13T11:51:08.488316Z","iopub.status.idle":"2022-02-13T11:51:08.747088Z","shell.execute_reply.started":"2022-02-13T11:51:08.488271Z","shell.execute_reply":"2022-02-13T11:51:08.745757Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["char_vocab_size = len(CHAR.vocab)\n","word_vocab_size = len(WORD.vocab)\n","emb = bert.embeddings.word_embeddings.weight.data"],"metadata":{"id":"hu5FfkVpzM4d","execution":{"iopub.status.busy":"2022-02-13T11:51:08.748730Z","iopub.execute_input":"2022-02-13T11:51:08.749084Z","iopub.status.idle":"2022-02-13T11:51:08.756442Z","shell.execute_reply.started":"2022-02-13T11:51:08.749033Z","shell.execute_reply":"2022-02-13T11:51:08.755401Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["  \n","n_epochs = 50\n","num_steps = len(train_iter) * n_epochs\n","args.dropout = 0.5\n","model = BiDAF2(args, pretrained=emb).to(device)\n","\n","no_decay = ['bias', 'LayerNorm.weight']\n","decay = 0.01\n","optimizer_grouped_parameters = [\n","    {'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': decay},\n","    {'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}\n","]\n","optimizer = ppb.AdamW(optimizer_grouped_parameters, lr=5e-5, eps=1e-8)\n","\n","criterion = nn.BCELoss()\n","\n","train_epochs (model, train_iter, val_iter, optimizer, criterion, scheduler=None, \n","              CLIP=1, N_EPOCHS=n_epochs, patience=5) \n","\n","print(test(model, weights='best-val-model.pt', iterator=test_iter))"],"metadata":{"id":"WI0c-M7ZzM4d","outputId":"dcf5e6f8-1080-4361-e099-5b34a8c1193d","execution":{"iopub.status.busy":"2022-02-13T11:51:08.758219Z","iopub.execute_input":"2022-02-13T11:51:08.758901Z","iopub.status.idle":"2022-02-13T12:54:50.507562Z","shell.execute_reply.started":"2022-02-13T11:51:08.758850Z","shell.execute_reply":"2022-02-13T12:54:50.506412Z"},"trusted":true},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 864x576 with 2 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"name":"stdout","text":"(0.6654471544715447, 0.7501050981538789)\n","output_type":"stream"}]},{"cell_type":"markdown","source":["Результат - 0.6654"],"metadata":{"id":"K36oiB3-Le-V"}},{"cell_type":"code","source":["predict (model, test_questions, test_passages, test_answers, weights='best-val-model.pt')"],"metadata":{"execution":{"iopub.status.busy":"2022-02-13T12:59:21.836787Z","iopub.execute_input":"2022-02-13T12:59:21.837090Z","iopub.status.idle":"2022-02-13T12:59:22.472646Z","shell.execute_reply.started":"2022-02-13T12:59:21.837057Z","shell.execute_reply":"2022-02-13T12:59:22.471734Z"},"trusted":true,"id":"BSj3PiUlLe-V","outputId":"bf80e9a0-1dd5-48f7-cb20-98e05137a82d"},"execution_count":null,"outputs":[{"name":"stdout","text":"Q.:  is manic depression the same as bi polar\nP.:  Bipolar disorder, previously known as manic depression, is a mental disorder that causes periods of depression and periods of abnormally elevated mood. The elevated mood is significant and is known as mania or hypomania, depending on its severity, or whether symptoms of psychosis are present. During mania, an individual behaves or feels abnormally energetic, happy, or irritable. Individuals often make poorly thought out decisions with little regard to the consequences. The need for sleep is usually reduced during manic phases. During periods of depression, there may be crying, a negative outlook on life, and poor eye contact with others. The risk of suicide among those with the illness is high at greater than 6 percent over 20 years, while self-harm occurs in 30--40 percent. Other mental health issues such as anxiety disorders and substance use disorder are commonly associated.\nNO\nQ.:  was whiskey galore based on a true story\nP.:  SS Politician was an 8000-ton cargo ship owned by T & J Harrison of Liverpool. It left Liverpool on 3 February 1941, bound for Kingston, Jamaica and New Orleans with a cargo including 28,000 cases of malt whisky. The ship sank off the north coast of Eriskay in the Outer Hebrides, off the west coast of Scotland, and much of the wreck's cargo was salvaged by the island's inhabitants. The story of the wreck and looting was the basis for the book and film Whisky Galore!.\nNO\nQ.:  are there plants on the international space station\nP.:  Plant research continued on the International Space Station. Biomass Production System was used on the ISS Expedition 4. The Vegetable Production System (Veggie) system was later used aboard ISS. Plants tested in Veggie before going into space included lettuce, Swiss chard, radishes, Chinese cabbage and peas. Red Romaine lettuce was grown in space on Expedition 40 which were harvested when mature, frozen and tested back on Earth. Expedition 44 members became the first American astronauts to eat plants grown in space on 10 August 2015, when their crop of Red Romaine was harvested. Since 2003 Russian cosmonauts have been eating half of their crop while the other half goes towards further research. In 2012, a sunflower bloomed aboard the ISS under the care of NASA astronaut Donald Pettit. In January 2016, US astronauts announced that a zinnia had blossomed aboard the ISS.\nYES\nQ.:  does the hockey puck have to cross the line to be a goal\nP.:  In ice hockey, a goal is scored when the puck entirely crosses the goal line between the two goal posts and below the goal crossbar. A goal awards one point to the team attacking the goal scored upon, regardless of which team the player who actually deflected the puck into the goal belongs to (see also own goal). Typically, a player on the team attempting to score shoots the puck with his/her stick towards the goal net opening, and a player on the opposing team called a goaltender tries to block the shot to prevent a goal from being scored against his/her team.\nNO\nQ.:  will there be a season 5 of shadowhunters\nP.:  In April 2017, it was announced that the series had been renewed for a third season of 20 episodes. The first half of ten episodes premiered on March 20, 2018. In June 2018, Freeform canceled the series after three seasons, but ordered two extra episodes to properly conclude the series' story; the second half of the third season is set to air in early 2019. As of May 15, 2018, 43 episodes of Shadowhunters have aired, concluding the first half of the third season.\nNO\n","output_type":"stream"}]},{"cell_type":"markdown","source":["При том, что метрика accuracy улучшилась, результаты на конкретном примере хуже, чем при использовании word_tokenizer."],"metadata":{"id":"D9bp9PKBQRTh"}},{"cell_type":"markdown","source":["## Часть 6. Итоги (2 балла)\n","Напишите краткое резюме проделанной работы. Сравните результаты всех разработанных моделей. Что помогло вам в выполнении работы, чего не хватало?"],"metadata":{"id":"IMXmEGfgqj4W"}},{"cell_type":"markdown","source":["Рассмотренные в рамках проекта подходы дали следующие показатели метрики accuracy:"],"metadata":{"id":"EGllmwZLhdKI"}},{"cell_type":"markdown","source":["\n","\n","*   Константное решение - 0.62\n","*   Использование fasttext на склейках вопросов и абзацев - 0.63\n","*   Обучение логистической регрессии на конкатенированных BERT-эмбеддингах вопроса и абзаца - 0.67\n","*   Файнтьюнинг предобученного BERTа с добавленными полносвязными слоями классификатора на конкатенации вопроса и предложения - 0.72 \n","*   Адаптация модели BiDAF для бинарного QA - 0.67\n","\n","\n","\n","\n","\n","\n","\n","\n"],"metadata":{"id":"g1JDXv1GiK-L"}},{"cell_type":"markdown","source":["Лучший результат дал файнтьюнинг предобученного BERTа с добавленными полносвязными слоями классификатора на конкатенации вопроса и предложения."],"metadata":{"id":"kOi_cUArjqFW"}},{"cell_type":"markdown","source":["В выполнении работы помогли наличие предобученных трансформеров и модели, предназначенные для развернутого (небинарного) QA. Однако, эксперименты показали, что задача бинарного QA существенно отличается от небинарного варианта (SQuAD и т.п.). Авторы датасета BoolQ в своей статье сделали аналогичный вывод. Беглый поиск подходов к решению задачи QA на BoolQ показал, что на данный момент хороших результатов в этой задаче никто не достиг."],"metadata":{"id":"yeFo2_j9j1g_"}},{"cell_type":"markdown","source":["Для выполнения проекта не хватило ресурсов бесплатных GPU-платформ (Colab, Kaggle), требуется существенно больший объем памяти GPU для использования батчей  большего размера и отказоустойчивости при проведении экспериментов. Также размер датасета BoolQ не позволяет полноценно обучать большие модели, требуется больше данных. Возможно, здесь помогла бы аугментация данных, но мне не удалось найти подходящий вариант аугментации."],"metadata":{"id":"z9Xs_PNbl4hl"}},{"cell_type":"code","source":[""],"metadata":{"id":"ItwZma1JzM4e"},"execution_count":null,"outputs":[]}]}