03-finetune-roberta.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "7e3e1a57-40d7-44d2-89f5-d59512d58998",
   "metadata": {},
   "source": [
    "# Fine-tune `roberta-base` model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "88b09ca9-2ea2-4704-90e5-7ab848bee52d",
   "metadata": {},
   "source": [
    "In this notebook, we will fine-tune [the `roberta-base` model][1], so that it can represent math-specific tokens.\n",
    "\n",
    " [1]: https://huggingface.co/roberta-base"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "103236fd-52e4-4128-8bd5-1dd8175ef378",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "docker.apollo.fi.muni.cz\n"
     ]
    }
   ],
   "source": [
    "! hostname"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "df0c9ef6-1117-491e-b71e-25ee6d0835b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "! pip install .[transformers,evaluation]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "60bbc1f6-ede7-41c8-84de-4ee0ae28e93a",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "! make tuned-roberta-base-text+latex"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc7e3e2a-a81a-4ed9-a310-2bc79e43fdb1",
   "metadata": {},
   "source": [
    "We will push the fine-tuned `roberta-base` model to [the 🤗 Model Hub][1].\n",
    "\n",
    " [1]: https://huggingface.co/witiko/mathberta"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "91307a74-7998-4939-9a01-a021a2fd8f72",
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoModel\n",
    "\n",
    "model = AutoModel.from_pretrained('./tuned-roberta-base-text+latex/MaskedLanguageModeling/')\n",
    "model.push_to_hub('mathberta')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d415b86c-ce34-47c7-a00c-e7f692de2178",
   "metadata": {},
   "source": [
    "## Intrinsic evaluation\n",
    "\n",
    "To validate our model during the training, we will plot the validation loss."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "865f3936-56f2-40cf-bda8-e5163e92f11c",
   "metadata": {},
   "source": [
    "### In-domain evaluation\n",
    "\n",
    "To see how well our model generalizes to unseen data from our target domain (scientific texts), we use a held-out dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "11aa01be-cda2-4079-8f8f-628e6c295086",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 1008 1011 52G May  5 01:52 dataset-text+latex.txt\n"
     ]
    }
   ],
   "source": [
    "%ls -lh dataset-text+latex.txt  # this is the training dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "d34506ca-1f91-4a3c-8b0a-194355218651",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 1008 1011 457M May  5 22:11 dataset-text+latex-validation.txt\n"
     ]
    }
   ],
   "source": [
    "%ls -lh dataset-text+latex-validation.txt  # this is the validation dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "de25ed39-1069-4215-88d3-0ca17296b765",
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "import json\n",
    "from pathlib import Path\n",
    "from statistics import mean\n",
    "from typing import List, Tuple\n",
    "\n",
    "def get_checkpoint_step(input_checkpoint_directory: Path) -> int:\n",
    "    _, checkpoint_step = input_checkpoint_directory.name.split('-')\n",
    "    checkpoint_step = int(checkpoint_step)\n",
    "    return checkpoint_step\n",
    "\n",
    "checkpoints = Path('tuned-roberta-base-text+latex.MLM-objective').glob('*/')\n",
    "latest_checkpoint = max(checkpoints, key=get_checkpoint_step)\n",
    "with (latest_checkpoint / 'trainer_state.json').open('rt') as f:\n",
    "    id_trainer_state = json.load(f)\n",
    "\n",
    "learning_rates = defaultdict(lambda: list())\n",
    "id_train_losses = defaultdict(lambda: list())\n",
    "id_validation_losses = defaultdict(lambda: list())\n",
    "\n",
    "for log in id_trainer_state['log_history']:\n",
    "    epoch = float(log['epoch'])\n",
    "    if 'learning_rate' in log:\n",
    "        learning_rate = float(log['learning_rate'])\n",
    "        learning_rates[epoch].append(learning_rate)\n",
    "    if 'loss' in log:\n",
    "        loss = float(log['loss'])\n",
    "        id_train_losses[epoch].append(loss)\n",
    "    if 'eval_loss' in log:\n",
    "        eval_loss = float(log['eval_loss'])\n",
    "        id_validation_losses[epoch].append(eval_loss)\n",
    "\n",
    "def average_values(item: Tuple[float, List[float]]) -> Tuple[float, float]:\n",
    "    epoch, values = item\n",
    "    average_value = mean(values)\n",
    "    return (epoch, average_value)\n",
    "\n",
    "learning_rates = sorted(map(average_values, learning_rates.items()))\n",
    "id_train_losses = sorted(map(average_values, id_train_losses.items()))\n",
    "id_validation_losses = sorted(map(average_values, id_validation_losses.items()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b59406c9-65f4-43d9-8b11-2867c9763250",
   "metadata": {},
   "source": [
    "### Out-of-domain evaluation\n",
    "\n",
    "To see how well our model generalizes to unseen data from a different domain, we use [a dataset of English texts from the European Constitution][1].\n",
    "\n",
    " [1]: https://opus.nlpl.eu/EUconst.php"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "38817277-9a57-40d1-a329-22d5f18a181e",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "! make out-of-domain-dataset-text-validation.json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "112f923b-20e2-4ea1-a572-303291a85e19",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 1008 1011 952K May 25 23:28 out-of-domain-dataset-text-validation.txt\n"
     ]
    }
   ],
   "source": [
    "%ls -lh out-of-domain-dataset-text-validation.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "32131cbd-05af-4577-96b1-1edcbecb7207",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "! make out-of-domain-dataset-text-validation.json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "9dc5052a-58eb-4a5f-afbc-574beff0576d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 1008 1011 12K Jun 11 21:01 out-of-domain-dataset-text-validation.json\n"
     ]
    }
   ],
   "source": [
    "%ls -lh out-of-domain-dataset-text-validation.json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c6ac7cbd-cc0b-4f7b-bf6d-f731f2188d1a",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('out-of-domain-dataset-text-validation.json', 'rt') as f:\n",
    "    ood_results = json.load(f)\n",
    "\n",
    "ood_validation_losses = defaultdict(lambda: list())\n",
    "\n",
    "for ood_result in ood_results:\n",
    "    epoch = ood_result['epoch']\n",
    "    loss = ood_result['loss']\n",
    "    ood_validation_losses[epoch].append(loss)\n",
    "\n",
    "ood_validation_losses = sorted(map(average_values, ood_validation_losses.items()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fc3472f-9650-4ea8-81ac-add10d34979f",
   "metadata": {},
   "source": [
    "### Results\n",
    "\n",
    "We will use the matplotlib library to plot the evaluation results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "6df7900f-e1c1-472a-b62f-31c452a0d0f9",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('fivethirtyeight')\n",
    "plt.rc('xtick', labelsize=16)\n",
    "plt.rc('ytick', labelsize=16)\n",
    "plt.rc('legend', fontsize=16)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "56603061-a5c5-4e69-93d6-d9a8bc0d1627",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 2))\n",
    "plt.plot(*zip(*learning_rates), label='learning rate')\n",
    "plt.xlabel('epoch')\n",
    "plt.xlim((0.0, epoch))\n",
    "_ = plt.ylabel('learning rate')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "154fa2f0-1632-41fb-8670-8f4c445aedf3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 5))\n",
    "plt.plot(*zip(*id_train_losses), label='training (in-domain)')\n",
    "plt.plot(*zip(*id_validation_losses), label='validation (in-domain)')\n",
    "plt.plot(*zip(*ood_validation_losses), label='validation (out-of-domain)')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('masked language modeling loss')\n",
    "plt.xlim((0.0, epoch))\n",
    "plt.ylim((0.75, 4.0))\n",
    "plt.legend()\n",
    "plt.savefig('paper/learning-curves.pdf', dpi=600)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}