notebooks/main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import itertools\n",
    "import json\n",
    "import operator\n",
    "import os\n",
    "from pathlib import Path\n",
    "from pprint import pprint\n",
    "import re\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as mpatches\n",
    "from matplotlib.lines import Line2D\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "from scipy import stats\n",
    "from tqdm.notebook import tqdm\n",
    "%matplotlib inline\n",
    "\n",
    "from IPython.display import set_matplotlib_formats\n",
    "set_matplotlib_formats('png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "toc-hr-collapsed": false
   },
   "source": [
    "## Load data and preprocess"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Map from test suite tag to high-level circuit.\n",
    "circuits = {\n",
    "    \"Licensing\": [\"npi\", \"reflexive\"],\n",
    "    \"Long-Distance Dependencies\": [\"fgd\", \"cleft\"],\n",
    "    \"Agreement\": [\"number\"],\n",
    "    \"Garden-Path Effects\": [\"npz\", \"mvrr\"],\n",
    "    \"Gross Syntactic State\": [\"subordination\"],\n",
    "    \"Center Embedding\": [\"center\"],\n",
    "}\n",
    "\n",
    "tag_to_circuit = {tag: circuit\n",
    "                  for circuit, tags in circuits.items()\n",
    "                  for tag in tags}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Map codenames to readable names for various columns.\n",
    "def format_pretrained(model_name):\n",
    "    return \"%s$^*$\" % model_name\n",
    "\n",
    "PRETTY_COLUMN_MAPS = [\n",
    "    (\"model_name\",\n",
    "     {\n",
    "        \"vanilla\": \"LSTM\",\n",
    "        \"ordered-neurons\": \"ON-LSTM\",\n",
    "        \"rnng\": \"RNNG\",\n",
    "        \"ngram\": \"n-gram\",\n",
    "        \"random\": \"Random\",\n",
    "         \n",
    "        \"gpt-2-pretrained\": format_pretrained(\"GPT-2\"),\n",
    "        \"gpt-2-xl-pretrained\": format_pretrained(\"GPT-2-XL\"),\n",
    "        \"gpt-2\": \"GPT-2\",\n",
    "        \"transformer-xl\": format_pretrained(\"Transformer-XL\"),\n",
    "        \"grnn\": format_pretrained(\"GRNN\"),\n",
    "        \"jrnn\": format_pretrained(\"JRNN\"),\n",
    "    }),\n",
    "    \n",
    "    (\"corpus\", lambda x: x.upper() if x else \"N/A\"),\n",
    "]\n",
    "PRETTY_COLUMNS = [\"pretty_%s\" % col for col, _ in PRETTY_COLUMN_MAPS]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Exclusions\n",
    "exclude_suite_re = re.compile(r\"^fgd-embed[34]|^gardenpath|^nn-nv\")\n",
    "exclude_models = [\"1gram\", \"ngram-no-rand\"] # \"ngram\", "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "ngram_models = [\"1gram\", \"ngram\", \"ngram-single\"]\n",
    "baseline_models = [\"random\"]\n",
    "\n",
    "# Models for which we designed a controlled training regime\n",
    "controlled_models = [\"ngram\", \"ordered-neurons\", \"vanilla\", \"rnng\", \"gpt-2\"]\n",
    "controlled_nonbpe_models = [\"ngram\", \"ordered-neurons\", \"vanilla\", \"rnng\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "ppl_data_path = Path(\"../data/raw/perplexity.csv\")\n",
    "test_suite_results_path = Path(\"../data/raw/sg_results\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tags missing circuit:  nn\n"
     ]
    }
   ],
   "source": [
    "perplexity_df = pd.read_csv(ppl_data_path, index_col=[\"model\", \"corpus\", \"seed\"])\n",
    "perplexity_df.index.set_names(\"model_name\", level=0, inplace=True)\n",
    "\n",
    "results_df = pd.concat([pd.read_csv(f) for f in test_suite_results_path.glob(\"*.csv\")])\n",
    "\n",
    "# Split model_id into constituent parts\n",
    "model_ids = results_df.model.str.split(\"_\", expand=True).rename(columns={0: \"model_name\", 1: \"corpus\", 2: \"seed\"})\n",
    "results_df = pd.concat([results_df, model_ids], axis=1).drop(columns=[\"model\"])\n",
    "results_df[\"seed\"] = results_df.seed.fillna(\"0\").astype(int)\n",
    "\n",
    "# Add tags\n",
    "results_df[\"tag\"] = results_df.suite.transform(lambda s: re.split(r\"[-_0-9]\", s)[0])\n",
    "results_df[\"circuit\"] = results_df.tag.map(tag_to_circuit)\n",
    "tags_missing_circuit = set(results_df.tag.unique()) - set(tag_to_circuit.keys())\n",
    "if tags_missing_circuit:\n",
    "    print(\"Tags missing circuit: \", \", \".join(tags_missing_circuit))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dropping 53922 results / 3 suites due to exclusions:\n",
      "fgd-embed3 fgd-embed4 nn-nv-rpl\n",
      "Dropping 6392 results due to dropping models: ['ngram-no-rand', '1gram']\n"
     ]
    }
   ],
   "source": [
    "# Exclude test suites\n",
    "exclude_filter = results_df.suite.str.contains(exclude_suite_re)\n",
    "print(\"Dropping %i results / %i suites due to exclusions:\"\n",
    "      % (exclude_filter.sum(), len(results_df[exclude_filter].suite.unique())))\n",
    "print(\" \".join(results_df[exclude_filter].suite.unique()))\n",
    "results_df = results_df[~exclude_filter]\n",
    "\n",
    "# Exclude models\n",
    "exclude_filter = results_df.model_name.isin(exclude_models)\n",
    "print(\"Dropping %i results due to dropping models:\" % exclude_filter.sum(), list(results_df[exclude_filter].model_name.unique()))\n",
    "results_df = results_df[~exclude_filter]\n",
    "\n",
    "# Exclude word-level controlled models with BPE tokenization\n",
    "exclude_filter = (results_df.model_name.isin(controlled_nonbpe_models)) & (results_df.corpus.str.endswith(\"bpe\"))\n",
    "results_df = results_df[~exclude_filter]\n",
    "\n",
    "# Exclude GPT-2 with word-level or SentencePieceBPE tokenization\n",
    "exclude_filter = ((results_df.model_name==\"gpt-2\") & ~(results_df.corpus.str.endswith(\"gptbpe\")))\n",
    "results_df = results_df[~exclude_filter]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Average across seeds of each ngram model.\n",
    "# The only difference between \"seeds\" of these model types are random differences in tie-breaking decisions.\n",
    "for ngram_model in ngram_models:\n",
    "    # Create a synthetic results_df with one ngram model, where each item is correct if more than half of\n",
    "    # the ngram seeds vote.\n",
    "    ngram_results_df = (results_df[results_df.model_name == ngram_model].copy()\n",
    "                        .groupby([\"model_name\", \"corpus\", \"suite\", \"item\", \"tag\", \"circuit\"])\n",
    "                        .agg({\"correct\": \"mean\"}) > 0.5).reset_index()\n",
    "    ngram_results_df[\"seed\"] = 0\n",
    "    \n",
    "    # Drop existing model results.\n",
    "    results_df = pd.concat([results_df[~(results_df.model_name == ngram_model)],\n",
    "                            ngram_results_df], sort=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prettify name columns, which we'll carry through data manipulations\n",
    "for column, map_fn in PRETTY_COLUMN_MAPS:\n",
    "    pretty_column = \"pretty_%s\" % column\n",
    "    results_df[pretty_column] = results_df[column].map(map_fn)\n",
    "    if results_df[pretty_column].isna().any():\n",
    "        print(\"WARNING: In prettifying %s, yielded NaN values:\" % column)\n",
    "        print(results_df[results_df[pretty_column].isna()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data prep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "suites_df = results_df.groupby([\"model_name\", \"corpus\", \"seed\", \"suite\"] + PRETTY_COLUMNS).correct.mean().reset_index()\n",
    "suites_df[\"tag\"] = suites_df.suite.transform(lambda s: re.split(r\"[-_0-9]\", s)[0])\n",
    "suites_df[\"circuit\"] = suites_df.tag.map(tag_to_circuit)\n",
    "\n",
    "# For controlled evaluation:\n",
    "# Compute a model's test suite accuracy relative to the mean accuracy on this test suite.\n",
    "# Only compute this on controlled models.\n",
    "def get_controlled_mean(suite_results):\n",
    "    # When computing test suite mean, first collapse test suite accuracies within model--corpus, then combine resulting means.\n",
    "    return suite_results[suite_results.model_name.isin(controlled_models)].groupby([\"model_name\", \"corpus\"]).correct.mean().mean()\n",
    "suite_means = suites_df.groupby(\"suite\").apply(get_controlled_mean)\n",
    "suites_df[\"correct_delta\"] = suites_df.apply(lambda r: r.correct - suite_means.loc[r.suite] if r.model_name in controlled_models else None, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We'll save this data to a CSV file for access from R, where we do\n",
    "# linear mixed-effects regression modeling.\n",
    "suites_df.to_csv(\"../data/suites_df.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Join PPL and accuracy data.\n",
    "joined_data = suites_df.groupby([\"model_name\", \"corpus\", \"seed\"] + PRETTY_COLUMNS)[[\"correct\", \"correct_delta\"]].agg(\"mean\")\n",
    "joined_data = pd.DataFrame(joined_data).join(perplexity_df).reset_index()\n",
    "joined_data.head()\n",
    "\n",
    "# Track BPE + size separately.\n",
    "joined_data[\"corpus_size\"] = joined_data.corpus.str.split(\"-\").apply(lambda tokens: tokens[1] if len(tokens) >= 2 else None)\n",
    "joined_data[\"corpus_bpe\"] = joined_data.corpus.str.split(\"-\").apply(lambda tokens: tokens[2] if len(tokens) > 2 else (\"none\" if len(tokens) >= 2 else None))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>circuit</th>\n",
       "      <th>pretty_model_name</th>\n",
       "      <th>pretty_corpus</th>\n",
       "      <th>correct</th>\n",
       "      <th>correct_delta</th>\n",
       "      <th>pid</th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_ppl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>Agreement</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.701754</td>\n",
       "      <td>0.455994</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>32.138346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>Center Embedding</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.910714</td>\n",
       "      <td>0.286756</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>32.138346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.855159</td>\n",
       "      <td>0.247057</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>32.138346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>Gross Syntactic State</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.967391</td>\n",
       "      <td>0.388496</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>32.138346</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>Licensing</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.586842</td>\n",
       "      <td>0.410417</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>32.138346</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  model_name           corpus        seed                circuit  \\\n",
       "0      gpt-2  bllip-lg-gptbpe  1581955288              Agreement   \n",
       "1      gpt-2  bllip-lg-gptbpe  1581955288       Center Embedding   \n",
       "2      gpt-2  bllip-lg-gptbpe  1581955288    Garden-Path Effects   \n",
       "3      gpt-2  bllip-lg-gptbpe  1581955288  Gross Syntactic State   \n",
       "4      gpt-2  bllip-lg-gptbpe  1581955288              Licensing   \n",
       "\n",
       "  pretty_model_name    pretty_corpus   correct  correct_delta  pid  test_loss  \\\n",
       "0             GPT-2  BLLIP-LG-GPTBPE  0.701754       0.455994  NaN        NaN   \n",
       "1             GPT-2  BLLIP-LG-GPTBPE  0.910714       0.286756  NaN        NaN   \n",
       "2             GPT-2  BLLIP-LG-GPTBPE  0.855159       0.247057  NaN        NaN   \n",
       "3             GPT-2  BLLIP-LG-GPTBPE  0.967391       0.388496  NaN        NaN   \n",
       "4             GPT-2  BLLIP-LG-GPTBPE  0.586842       0.410417  NaN        NaN   \n",
       "\n",
       "    test_ppl  \n",
       "0  32.138346  \n",
       "1  32.138346  \n",
       "2  32.138346  \n",
       "3  32.138346  \n",
       "4  32.138346  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Join PPL and accuracy data, splitting on circuit.\n",
    "joined_data_circuits = suites_df.groupby([\"model_name\", \"corpus\", \"seed\", \"circuit\"] + PRETTY_COLUMNS)[[\"correct\", \"correct_delta\"]].agg(\"mean\")\n",
    "joined_data_circuits = pd.DataFrame(joined_data_circuits).reset_index().set_index([\"model_name\", \"corpus\", \"seed\"]).join(perplexity_df).reset_index()\n",
    "joined_data_circuits.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>suite</th>\n",
       "      <th>pretty_model_name</th>\n",
       "      <th>pretty_corpus</th>\n",
       "      <th>correct</th>\n",
       "      <th>tag</th>\n",
       "      <th>circuit</th>\n",
       "      <th>correct_delta</th>\n",
       "      <th>has_modifier</th>\n",
       "      <th>test_suite_base</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>center_embed</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.892857</td>\n",
       "      <td>center</td>\n",
       "      <td>Center Embedding</td>\n",
       "      <td>0.231399</td>\n",
       "      <td>False</td>\n",
       "      <td>center_emb</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>center_embed_mod</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.928571</td>\n",
       "      <td>center</td>\n",
       "      <td>Center Embedding</td>\n",
       "      <td>0.342113</td>\n",
       "      <td>True</td>\n",
       "      <td>center_emb</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>cleft</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>cleft</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "      <td>0.211458</td>\n",
       "      <td>False</td>\n",
       "      <td>cleft</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>cleft_modifier</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.775000</td>\n",
       "      <td>cleft</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "      <td>0.287083</td>\n",
       "      <td>True</td>\n",
       "      <td>cleft</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>gpt-2</td>\n",
       "      <td>bllip-lg-gptbpe</td>\n",
       "      <td>1581955288</td>\n",
       "      <td>mvrr</td>\n",
       "      <td>GPT-2</td>\n",
       "      <td>BLLIP-LG-GPTBPE</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>mvrr</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>0.148958</td>\n",
       "      <td>False</td>\n",
       "      <td>v</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  model_name           corpus        seed             suite pretty_model_name  \\\n",
       "0      gpt-2  bllip-lg-gptbpe  1581955288      center_embed             GPT-2   \n",
       "1      gpt-2  bllip-lg-gptbpe  1581955288  center_embed_mod             GPT-2   \n",
       "2      gpt-2  bllip-lg-gptbpe  1581955288             cleft             GPT-2   \n",
       "3      gpt-2  bllip-lg-gptbpe  1581955288    cleft_modifier             GPT-2   \n",
       "8      gpt-2  bllip-lg-gptbpe  1581955288              mvrr             GPT-2   \n",
       "\n",
       "     pretty_corpus   correct     tag                     circuit  \\\n",
       "0  BLLIP-LG-GPTBPE  0.892857  center            Center Embedding   \n",
       "1  BLLIP-LG-GPTBPE  0.928571  center            Center Embedding   \n",
       "2  BLLIP-LG-GPTBPE  1.000000   cleft  Long-Distance Dependencies   \n",
       "3  BLLIP-LG-GPTBPE  0.775000   cleft  Long-Distance Dependencies   \n",
       "8  BLLIP-LG-GPTBPE  0.714286    mvrr         Garden-Path Effects   \n",
       "\n",
       "   correct_delta  has_modifier test_suite_base  \n",
       "0       0.231399         False      center_emb  \n",
       "1       0.342113          True      center_emb  \n",
       "2       0.211458         False           cleft  \n",
       "3       0.287083          True           cleft  \n",
       "8       0.148958         False               v  "
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Analyze stability to modification.\n",
    "def has_modifier(ts):\n",
    "    if ts.endswith((\"_modifier\", \"_mod\")):\n",
    "        return True\n",
    "    else:\n",
    "        return None\n",
    "suites_df[\"has_modifier\"] = suites_df.suite.transform(has_modifier)\n",
    "\n",
    "# Mark \"non-modifier\" test suites\n",
    "modifier_ts = suites_df[suites_df.has_modifier == True].suite.unique()\n",
    "no_modifier_ts = [re.sub(r\"_mod(ifier)?$\", \"\", ts) for ts in modifier_ts]\n",
    "suites_df.loc[suites_df.suite.isin(no_modifier_ts), \"has_modifier\"] = False\n",
    "# Store subset of test suites which have definite modifier/no-modifier marking\n",
    "suites_df_mod = suites_df[~(suites_df.has_modifier.isna())].copy()\n",
    "suites_df_mod[\"has_modifier\"] = suites_df_mod.has_modifier.astype(bool)\n",
    "# Get base test suite (without modifier/no-modifier marking)\n",
    "suites_df_mod[\"test_suite_base\"] = suites_df_mod.suite.transform(lambda ts: ts.strip(\"_no-modifier\").strip(\"_modifier\"))\n",
    "suites_df_mod.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Checks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Missing perplexity results for:\n",
      "{('gpt-2-pretrained', '', 0),\n",
      " ('gpt-2-xl-pretrained', '', 0),\n",
      " ('grnn', '', 0),\n",
      " ('jrnn', '', 0),\n",
      " ('random', '', 1),\n",
      " ('random', '', 2),\n",
      " ('random', '', 3),\n",
      " ('random', '', 4),\n",
      " ('random', '', 5),\n",
      " ('random', '', 6),\n",
      " ('random', '', 7),\n",
      " ('random', '', 8),\n",
      " ('random', '', 9),\n",
      " ('random', '', 10),\n",
      " ('random', '', 11),\n",
      " ('random', '', 12),\n",
      " ('random', '', 13),\n",
      " ('random', '', 14),\n",
      " ('random', '', 15),\n",
      " ('random', '', 16),\n",
      " ('random', '', 17),\n",
      " ('random', '', 18),\n",
      " ('random', '', 19),\n",
      " ('random', '', 20),\n",
      " ('random', '', 21),\n",
      " ('random', '', 22),\n",
      " ('random', '', 23),\n",
      " ('random', '', 24),\n",
      " ('random', '', 25),\n",
      " ('random', '', 26),\n",
      " ('random', '', 27),\n",
      " ('random', '', 28),\n",
      " ('random', '', 29),\n",
      " ('random', '', 30),\n",
      " ('random', '', 31),\n",
      " ('random', '', 32),\n",
      " ('random', '', 33),\n",
      " ('random', '', 34),\n",
      " ('random', '', 35),\n",
      " ('random', '', 36),\n",
      " ('random', '', 37),\n",
      " ('random', '', 38),\n",
      " ('random', '', 39),\n",
      " ('random', '', 40),\n",
      " ('random', '', 41),\n",
      " ('random', '', 42),\n",
      " ('random', '', 43),\n",
      " ('random', '', 44),\n",
      " ('random', '', 45),\n",
      " ('random', '', 46),\n",
      " ('random', '', 47),\n",
      " ('random', '', 48),\n",
      " ('random', '', 49),\n",
      " ('random', '', 50),\n",
      " ('random', '', 51),\n",
      " ('random', '', 52),\n",
      " ('random', '', 53),\n",
      " ('random', '', 54),\n",
      " ('random', '', 55),\n",
      " ('random', '', 56),\n",
      " ('random', '', 57),\n",
      " ('random', '', 58),\n",
      " ('random', '', 59),\n",
      " ('random', '', 60),\n",
      " ('random', '', 61),\n",
      " ('random', '', 62),\n",
      " ('random', '', 63),\n",
      " ('random', '', 64),\n",
      " ('random', '', 65),\n",
      " ('random', '', 66),\n",
      " ('random', '', 67),\n",
      " ('random', '', 68),\n",
      " ('random', '', 69),\n",
      " ('random', '', 70),\n",
      " ('random', '', 71),\n",
      " ('random', '', 72),\n",
      " ('random', '', 73),\n",
      " ('random', '', 74),\n",
      " ('random', '', 75),\n",
      " ('random', '', 76),\n",
      " ('random', '', 77),\n",
      " ('random', '', 78),\n",
      " ('random', '', 79),\n",
      " ('random', '', 80),\n",
      " ('random', '', 81),\n",
      " ('random', '', 82),\n",
      " ('random', '', 83),\n",
      " ('random', '', 84),\n",
      " ('random', '', 85),\n",
      " ('random', '', 86),\n",
      " ('random', '', 87),\n",
      " ('random', '', 88),\n",
      " ('random', '', 89),\n",
      " ('random', '', 90),\n",
      " ('random', '', 91),\n",
      " ('random', '', 92),\n",
      " ('random', '', 93),\n",
      " ('random', '', 94),\n",
      " ('random', '', 95),\n",
      " ('random', '', 96),\n",
      " ('random', '', 97),\n",
      " ('random', '', 98),\n",
      " ('random', '', 99),\n",
      " ('random', '', 100),\n",
      " ('random', '', 101),\n",
      " ('random', '', 102),\n",
      " ('random', '', 103),\n",
      " ('random', '', 104),\n",
      " ('random', '', 105),\n",
      " ('random', '', 106),\n",
      " ('random', '', 107),\n",
      " ('random', '', 108),\n",
      " ('random', '', 109),\n",
      " ('random', '', 110),\n",
      " ('random', '', 111),\n",
      " ('random', '', 112),\n",
      " ('random', '', 113),\n",
      " ('random', '', 114),\n",
      " ('random', '', 115),\n",
      " ('random', '', 116),\n",
      " ('random', '', 117),\n",
      " ('random', '', 118),\n",
      " ('random', '', 119),\n",
      " ('random', '', 120),\n",
      " ('random', '', 121),\n",
      " ('random', '', 122),\n",
      " ('random', '', 123),\n",
      " ('random', '', 124),\n",
      " ('random', '', 125),\n",
      " ('random', '', 126),\n",
      " ('random', '', 127),\n",
      " ('random', '', 128),\n",
      " ('random', '', 129),\n",
      " ('random', '', 130),\n",
      " ('random', '', 131),\n",
      " ('random', '', 132),\n",
      " ('random', '', 133),\n",
      " ('random', '', 134),\n",
      " ('random', '', 135),\n",
      " ('random', '', 136),\n",
      " ('random', '', 137),\n",
      " ('random', '', 138),\n",
      " ('random', '', 139),\n",
      " ('random', '', 140),\n",
      " ('random', '', 141),\n",
      " ('random', '', 142),\n",
      " ('random', '', 143),\n",
      " ('random', '', 144),\n",
      " ('random', '', 145),\n",
      " ('random', '', 146),\n",
      " ('random', '', 147),\n",
      " ('random', '', 148),\n",
      " ('random', '', 149),\n",
      " ('random', '', 150),\n",
      " ('random', '', 151),\n",
      " ('random', '', 152),\n",
      " ('random', '', 153),\n",
      " ('random', '', 154),\n",
      " ('random', '', 155),\n",
      " ('random', '', 156),\n",
      " ('random', '', 157),\n",
      " ('random', '', 158),\n",
      " ('random', '', 159),\n",
      " ('random', '', 160),\n",
      " ('random', '', 161),\n",
      " ('random', '', 162),\n",
      " ('random', '', 163),\n",
      " ('random', '', 164),\n",
      " ('random', '', 165),\n",
      " ('random', '', 166),\n",
      " ('random', '', 167),\n",
      " ('random', '', 168),\n",
      " ('random', '', 169),\n",
      " ('random', '', 170),\n",
      " ('random', '', 171),\n",
      " ('random', '', 172),\n",
      " ('random', '', 173),\n",
      " ('random', '', 174),\n",
      " ('random', '', 175),\n",
      " ('random', '', 176),\n",
      " ('random', '', 177),\n",
      " ('random', '', 178),\n",
      " ('random', '', 179),\n",
      " ('random', '', 180),\n",
      " ('random', '', 181),\n",
      " ('random', '', 182),\n",
      " ('random', '', 183),\n",
      " ('random', '', 184),\n",
      " ('random', '', 185),\n",
      " ('random', '', 186),\n",
      " ('random', '', 187),\n",
      " ('random', '', 188),\n",
      " ('random', '', 189),\n",
      " ('random', '', 190),\n",
      " ('random', '', 191),\n",
      " ('random', '', 192),\n",
      " ('random', '', 193),\n",
      " ('random', '', 194),\n",
      " ('random', '', 195),\n",
      " ('random', '', 196),\n",
      " ('random', '', 197),\n",
      " ('random', '', 198),\n",
      " ('random', '', 199),\n",
      " ('random', '', 200),\n",
      " ('random', '', 201),\n",
      " ('random', '', 202),\n",
      " ('random', '', 203),\n",
      " ('random', '', 204),\n",
      " ('random', '', 205),\n",
      " ('random', '', 206),\n",
      " ('random', '', 207),\n",
      " ('random', '', 208),\n",
      " ('random', '', 209),\n",
      " ('random', '', 210),\n",
      " ('random', '', 211),\n",
      " ('random', '', 212),\n",
      " ('random', '', 213),\n",
      " ('random', '', 214),\n",
      " ('random', '', 215),\n",
      " ('random', '', 216),\n",
      " ('random', '', 217),\n",
      " ('random', '', 218),\n",
      " ('random', '', 219),\n",
      " ('random', '', 220),\n",
      " ('random', '', 221),\n",
      " ('random', '', 222),\n",
      " ('random', '', 223),\n",
      " ('random', '', 224),\n",
      " ('random', '', 225),\n",
      " ('random', '', 226),\n",
      " ('random', '', 227),\n",
      " ('random', '', 228),\n",
      " ('random', '', 229),\n",
      " ('random', '', 230),\n",
      " ('random', '', 231),\n",
      " ('random', '', 232),\n",
      " ('random', '', 233),\n",
      " ('random', '', 234),\n",
      " ('random', '', 235),\n",
      " ('random', '', 236),\n",
      " ('random', '', 237),\n",
      " ('random', '', 238),\n",
      " ('random', '', 239),\n",
      " ('random', '', 240),\n",
      " ('random', '', 241),\n",
      " ('random', '', 242),\n",
      " ('random', '', 243),\n",
      " ('random', '', 244),\n",
      " ('random', '', 245),\n",
      " ('random', '', 246),\n",
      " ('random', '', 247),\n",
      " ('random', '', 248),\n",
      " ('random', '', 249),\n",
      " ('random', '', 250),\n",
      " ('random', '', 251),\n",
      " ('random', '', 252),\n",
      " ('random', '', 253),\n",
      " ('random', '', 254),\n",
      " ('random', '', 255),\n",
      " ('random', '', 256),\n",
      " ('random', '', 257),\n",
      " ('random', '', 258),\n",
      " ('random', '', 259),\n",
      " ('random', '', 260),\n",
      " ('random', '', 261),\n",
      " ('random', '', 262),\n",
      " ('random', '', 263),\n",
      " ('random', '', 264),\n",
      " ('random', '', 265),\n",
      " ('random', '', 266),\n",
      " ('random', '', 267),\n",
      " ('random', '', 268),\n",
      " ('random', '', 269),\n",
      " ('random', '', 270),\n",
      " ('random', '', 271),\n",
      " ('random', '', 272),\n",
      " ('random', '', 273),\n",
      " ('random', '', 274),\n",
      " ('random', '', 275),\n",
      " ('random', '', 276),\n",
      " ('random', '', 277),\n",
      " ('random', '', 278),\n",
      " ('random', '', 279),\n",
      " ('random', '', 280),\n",
      " ('random', '', 281),\n",
      " ('random', '', 282),\n",
      " ('random', '', 283),\n",
      " ('random', '', 284),\n",
      " ('random', '', 285),\n",
      " ('random', '', 286),\n",
      " ('random', '', 287),\n",
      " ('random', '', 288),\n",
      " ('random', '', 289),\n",
      " ('random', '', 290),\n",
      " ('random', '', 291),\n",
      " ('random', '', 292),\n",
      " ('random', '', 293),\n",
      " ('random', '', 294),\n",
      " ('random', '', 295),\n",
      " ('random', '', 296),\n",
      " ('random', '', 297),\n",
      " ('random', '', 298),\n",
      " ('random', '', 299),\n",
      " ('random', '', 300),\n",
      " ('random', '', 301),\n",
      " ('random', '', 302),\n",
      " ('random', '', 303),\n",
      " ('random', '', 304),\n",
      " ('random', '', 305),\n",
      " ('random', '', 306),\n",
      " ('random', '', 307),\n",
      " ('random', '', 308),\n",
      " ('random', '', 309),\n",
      " ('random', '', 310),\n",
      " ('random', '', 311),\n",
      " ('random', '', 312),\n",
      " ('random', '', 313),\n",
      " ('random', '', 314),\n",
      " ('random', '', 315),\n",
      " ('random', '', 316),\n",
      " ('random', '', 317),\n",
      " ('random', '', 318),\n",
      " ('random', '', 319),\n",
      " ('random', '', 320),\n",
      " ('random', '', 321),\n",
      " ('random', '', 322),\n",
      " ('random', '', 323),\n",
      " ('random', '', 324),\n",
      " ('random', '', 325),\n",
      " ('random', '', 326),\n",
      " ('random', '', 327),\n",
      " ('random', '', 328),\n",
      " ('random', '', 329),\n",
      " ('random', '', 330),\n",
      " ('random', '', 331),\n",
      " ('random', '', 332),\n",
      " ('random', '', 333),\n",
      " ('random', '', 334),\n",
      " ('random', '', 335),\n",
      " ('random', '', 336),\n",
      " ('random', '', 337),\n",
      " ('random', '', 338),\n",
      " ('random', '', 339),\n",
      " ('random', '', 340),\n",
      " ('random', '', 341),\n",
      " ('random', '', 342),\n",
      " ('random', '', 343),\n",
      " ('random', '', 344),\n",
      " ('random', '', 345),\n",
      " ('random', '', 346),\n",
      " ('random', '', 347),\n",
      " ('random', '', 348),\n",
      " ('random', '', 349),\n",
      " ('random', '', 350),\n",
      " ('random', '', 351),\n",
      " ('random', '', 352),\n",
      " ('random', '', 353),\n",
      " ('random', '', 354),\n",
      " ('random', '', 355),\n",
      " ('random', '', 356),\n",
      " ('random', '', 357),\n",
      " ('random', '', 358),\n",
      " ('random', '', 359),\n",
      " ('random', '', 360),\n",
      " ('random', '', 361),\n",
      " ('random', '', 362),\n",
      " ('random', '', 363),\n",
      " ('random', '', 364),\n",
      " ('random', '', 365),\n",
      " ('random', '', 366),\n",
      " ('random', '', 367),\n",
      " ('random', '', 368),\n",
      " ('random', '', 369),\n",
      " ('random', '', 370),\n",
      " ('random', '', 371),\n",
      " ('random', '', 372),\n",
      " ('random', '', 373),\n",
      " ('random', '', 374),\n",
      " ('random', '', 375),\n",
      " ('random', '', 376),\n",
      " ('random', '', 377),\n",
      " ('random', '', 378),\n",
      " ('random', '', 379),\n",
      " ('random', '', 380),\n",
      " ('random', '', 381),\n",
      " ('random', '', 382),\n",
      " ('random', '', 383),\n",
      " ('random', '', 384),\n",
      " ('random', '', 385),\n",
      " ('random', '', 386),\n",
      " ('random', '', 387),\n",
      " ('random', '', 388),\n",
      " ('random', '', 389),\n",
      " ('random', '', 390),\n",
      " ('random', '', 391),\n",
      " ('random', '', 392),\n",
      " ('random', '', 393),\n",
      " ('random', '', 394),\n",
      " ('random', '', 395),\n",
      " ('random', '', 396),\n",
      " ('random', '', 397),\n",
      " ('random', '', 398),\n",
      " ('random', '', 399),\n",
      " ('random', '', 400),\n",
      " ('random', '', 401),\n",
      " ('random', '', 402),\n",
      " ('random', '', 403),\n",
      " ('random', '', 404),\n",
      " ('random', '', 405),\n",
      " ('random', '', 406),\n",
      " ('random', '', 407),\n",
      " ('random', '', 408),\n",
      " ('random', '', 409),\n",
      " ('random', '', 410),\n",
      " ('random', '', 411),\n",
      " ('random', '', 412),\n",
      " ('random', '', 413),\n",
      " ('random', '', 414),\n",
      " ('random', '', 415),\n",
      " ('random', '', 416),\n",
      " ('random', '', 417),\n",
      " ('random', '', 418),\n",
      " ('random', '', 419),\n",
      " ('random', '', 420),\n",
      " ('random', '', 421),\n",
      " ('random', '', 422),\n",
      " ('random', '', 423),\n",
      " ('random', '', 424),\n",
      " ('random', '', 425),\n",
      " ('random', '', 426),\n",
      " ('random', '', 427),\n",
      " ('random', '', 428),\n",
      " ('random', '', 429),\n",
      " ('random', '', 430),\n",
      " ('random', '', 431),\n",
      " ('random', '', 432),\n",
      " ('random', '', 433),\n",
      " ('random', '', 434),\n",
      " ('random', '', 435),\n",
      " ('random', '', 436),\n",
      " ('random', '', 437),\n",
      " ('random', '', 438),\n",
      " ('random', '', 439),\n",
      " ('random', '', 440),\n",
      " ('random', '', 441),\n",
      " ('random', '', 442),\n",
      " ('random', '', 443),\n",
      " ('random', '', 444),\n",
      " ('random', '', 445),\n",
      " ('random', '', 446),\n",
      " ('random', '', 447),\n",
      " ('random', '', 448),\n",
      " ('random', '', 449),\n",
      " ('random', '', 450),\n",
      " ('random', '', 451),\n",
      " ('random', '', 452),\n",
      " ('random', '', 453),\n",
      " ('random', '', 454),\n",
      " ('random', '', 455),\n",
      " ('random', '', 456),\n",
      " ('random', '', 457),\n",
      " ('random', '', 458),\n",
      " ('random', '', 459),\n",
      " ('random', '', 460),\n",
      " ('random', '', 461),\n",
      " ('random', '', 462),\n",
      " ('random', '', 463),\n",
      " ('random', '', 464),\n",
      " ('random', '', 465),\n",
      " ('random', '', 466),\n",
      " ('random', '', 467),\n",
      " ('random', '', 468),\n",
      " ('random', '', 469),\n",
      " ('random', '', 470),\n",
      " ('random', '', 471),\n",
      " ('random', '', 472),\n",
      " ('random', '', 473),\n",
      " ('random', '', 474),\n",
      " ('random', '', 475),\n",
      " ('random', '', 476),\n",
      " ('random', '', 477),\n",
      " ('random', '', 478),\n",
      " ('random', '', 479),\n",
      " ('random', '', 480),\n",
      " ('random', '', 481),\n",
      " ('random', '', 482),\n",
      " ('random', '', 483),\n",
      " ('random', '', 484),\n",
      " ('random', '', 485),\n",
      " ('random', '', 486),\n",
      " ('random', '', 487),\n",
      " ('random', '', 488),\n",
      " ('random', '', 489),\n",
      " ('random', '', 490),\n",
      " ('random', '', 491),\n",
      " ('random', '', 492),\n",
      " ('random', '', 493),\n",
      " ('random', '', 494),\n",
      " ('random', '', 495),\n",
      " ('random', '', 496),\n",
      " ('random', '', 497),\n",
      " ('random', '', 498),\n",
      " ('random', '', 499),\n",
      " ('random', '', 500),\n",
      " ('random', '', 501),\n",
      " ('random', '', 502),\n",
      " ('random', '', 503),\n",
      " ('random', '', 504),\n",
      " ('random', '', 505),\n",
      " ('random', '', 506),\n",
      " ('random', '', 507),\n",
      " ('random', '', 508),\n",
      " ('random', '', 509),\n",
      " ('random', '', 510),\n",
      " ('random', '', 511),\n",
      " ('random', '', 512),\n",
      " ('random', '', 513),\n",
      " ('random', '', 514),\n",
      " ('random', '', 515),\n",
      " ('random', '', 516),\n",
      " ('random', '', 517),\n",
      " ('random', '', 518),\n",
      " ('random', '', 519),\n",
      " ('random', '', 520),\n",
      " ('random', '', 521),\n",
      " ('random', '', 522),\n",
      " ('random', '', 523),\n",
      " ('random', '', 524),\n",
      " ('random', '', 525),\n",
      " ('random', '', 526),\n",
      " ('random', '', 527),\n",
      " ('random', '', 528),\n",
      " ('random', '', 529),\n",
      " ('random', '', 530),\n",
      " ('random', '', 531),\n",
      " ('random', '', 532),\n",
      " ('random', '', 533),\n",
      " ('random', '', 534),\n",
      " ('random', '', 535),\n",
      " ('random', '', 536),\n",
      " ('random', '', 537),\n",
      " ('random', '', 538),\n",
      " ('random', '', 539),\n",
      " ('random', '', 540),\n",
      " ('random', '', 541),\n",
      " ('random', '', 542),\n",
      " ('random', '', 543),\n",
      " ('random', '', 544),\n",
      " ('random', '', 545),\n",
      " ('random', '', 546),\n",
      " ('random', '', 547),\n",
      " ('random', '', 548),\n",
      " ('random', '', 549),\n",
      " ('random', '', 550),\n",
      " ('random', '', 551),\n",
      " ('random', '', 552),\n",
      " ('random', '', 553),\n",
      " ('random', '', 554),\n",
      " ('random', '', 555),\n",
      " ('random', '', 556),\n",
      " ('random', '', 557),\n",
      " ('random', '', 558),\n",
      " ('random', '', 559),\n",
      " ('random', '', 560),\n",
      " ('random', '', 561),\n",
      " ('random', '', 562),\n",
      " ('random', '', 563),\n",
      " ('random', '', 564),\n",
      " ('random', '', 565),\n",
      " ('random', '', 566),\n",
      " ('random', '', 567),\n",
      " ('random', '', 568),\n",
      " ('random', '', 569),\n",
      " ('random', '', 570),\n",
      " ('random', '', 571),\n",
      " ('random', '', 572),\n",
      " ('random', '', 573),\n",
      " ('random', '', 574),\n",
      " ('random', '', 575),\n",
      " ('random', '', 576),\n",
      " ('random', '', 577),\n",
      " ('random', '', 578),\n",
      " ('random', '', 579),\n",
      " ('random', '', 580),\n",
      " ('random', '', 581),\n",
      " ('random', '', 582),\n",
      " ('random', '', 583),\n",
      " ('random', '', 584),\n",
      " ('random', '', 585),\n",
      " ('random', '', 586),\n",
      " ('random', '', 587),\n",
      " ('random', '', 588),\n",
      " ('random', '', 589),\n",
      " ('random', '', 590),\n",
      " ('random', '', 591),\n",
      " ('random', '', 592),\n",
      " ('random', '', 593),\n",
      " ('random', '', 594),\n",
      " ('random', '', 595),\n",
      " ('random', '', 596),\n",
      " ('random', '', 597),\n",
      " ('random', '', 598),\n",
      " ('random', '', 599),\n",
      " ('random', '', 600),\n",
      " ('random', '', 601),\n",
      " ('random', '', 602),\n",
      " ('random', '', 603),\n",
      " ('random', '', 604),\n",
      " ('random', '', 605),\n",
      " ('random', '', 606),\n",
      " ('random', '', 607),\n",
      " ('random', '', 608),\n",
      " ('random', '', 609),\n",
      " ('random', '', 610),\n",
      " ('random', '', 611),\n",
      " ('random', '', 612),\n",
      " ('random', '', 613),\n",
      " ('random', '', 614),\n",
      " ('random', '', 615),\n",
      " ('random', '', 616),\n",
      " ('random', '', 617),\n",
      " ('random', '', 618),\n",
      " ('random', '', 619),\n",
      " ('random', '', 620),\n",
      " ('random', '', 621),\n",
      " ('random', '', 622),\n",
      " ('random', '', 623),\n",
      " ('random', '', 624),\n",
      " ('random', '', 625),\n",
      " ('random', '', 626),\n",
      " ('random', '', 627),\n",
      " ('random', '', 628),\n",
      " ('random', '', 629),\n",
      " ('random', '', 630),\n",
      " ('random', '', 631),\n",
      " ('random', '', 632),\n",
      " ('random', '', 633),\n",
      " ('random', '', 634),\n",
      " ('random', '', 635),\n",
      " ('random', '', 636),\n",
      " ('random', '', 637),\n",
      " ('random', '', 638),\n",
      " ('random', '', 639),\n",
      " ('random', '', 640),\n",
      " ('random', '', 641),\n",
      " ('random', '', 642),\n",
      " ('random', '', 643),\n",
      " ('random', '', 644),\n",
      " ('random', '', 645),\n",
      " ('random', '', 646),\n",
      " ('random', '', 647),\n",
      " ('random', '', 648),\n",
      " ('random', '', 649),\n",
      " ('random', '', 650),\n",
      " ('random', '', 651),\n",
      " ('random', '', 652),\n",
      " ('random', '', 653),\n",
      " ('random', '', 654),\n",
      " ('random', '', 655),\n",
      " ('random', '', 656),\n",
      " ('random', '', 657),\n",
      " ('random', '', 658),\n",
      " ('random', '', 659),\n",
      " ('random', '', 660),\n",
      " ('random', '', 661),\n",
      " ('random', '', 662),\n",
      " ('random', '', 663),\n",
      " ('random', '', 664),\n",
      " ('random', '', 665),\n",
      " ('random', '', 666),\n",
      " ('random', '', 667),\n",
      " ('random', '', 668),\n",
      " ('random', '', 669),\n",
      " ('random', '', 670),\n",
      " ('random', '', 671),\n",
      " ('random', '', 672),\n",
      " ('random', '', 673),\n",
      " ('random', '', 674),\n",
      " ('random', '', 675),\n",
      " ('random', '', 676),\n",
      " ('random', '', 677),\n",
      " ('random', '', 678),\n",
      " ('random', '', 679),\n",
      " ('random', '', 680),\n",
      " ('random', '', 681),\n",
      " ('random', '', 682),\n",
      " ('random', '', 683),\n",
      " ('random', '', 684),\n",
      " ('random', '', 685),\n",
      " ('random', '', 686),\n",
      " ('random', '', 687),\n",
      " ('random', '', 688),\n",
      " ('random', '', 689),\n",
      " ('random', '', 690),\n",
      " ('random', '', 691),\n",
      " ('random', '', 692),\n",
      " ('random', '', 693),\n",
      " ('random', '', 694),\n",
      " ('random', '', 695),\n",
      " ('random', '', 696),\n",
      " ('random', '', 697),\n",
      " ('random', '', 698),\n",
      " ('random', '', 699),\n",
      " ('random', '', 700),\n",
      " ('random', '', 701),\n",
      " ('random', '', 702),\n",
      " ('random', '', 703),\n",
      " ('random', '', 704),\n",
      " ('random', '', 705),\n",
      " ('random', '', 706),\n",
      " ('random', '', 707),\n",
      " ('random', '', 708),\n",
      " ('random', '', 709),\n",
      " ('random', '', 710),\n",
      " ('random', '', 711),\n",
      " ('random', '', 712),\n",
      " ('random', '', 713),\n",
      " ('random', '', 714),\n",
      " ('random', '', 715),\n",
      " ('random', '', 716),\n",
      " ('random', '', 717),\n",
      " ('random', '', 718),\n",
      " ('random', '', 719),\n",
      " ('random', '', 720),\n",
      " ('random', '', 721),\n",
      " ('random', '', 722),\n",
      " ('random', '', 723),\n",
      " ('random', '', 724),\n",
      " ('random', '', 725),\n",
      " ('random', '', 726),\n",
      " ('random', '', 727),\n",
      " ('random', '', 728),\n",
      " ('random', '', 729),\n",
      " ('random', '', 730),\n",
      " ('random', '', 731),\n",
      " ('random', '', 732),\n",
      " ('random', '', 733),\n",
      " ('random', '', 734),\n",
      " ('random', '', 735),\n",
      " ('random', '', 736),\n",
      " ('random', '', 737),\n",
      " ('random', '', 738),\n",
      " ('random', '', 739),\n",
      " ('random', '', 740),\n",
      " ('random', '', 741),\n",
      " ('random', '', 742),\n",
      " ('random', '', 743),\n",
      " ('random', '', 744),\n",
      " ('random', '', 745),\n",
      " ('random', '', 746),\n",
      " ('random', '', 747),\n",
      " ('random', '', 748),\n",
      " ('random', '', 749),\n",
      " ('random', '', 750),\n",
      " ('random', '', 751),\n",
      " ('random', '', 752),\n",
      " ('random', '', 753),\n",
      " ('random', '', 754),\n",
      " ('random', '', 755),\n",
      " ('random', '', 756),\n",
      " ('random', '', 757),\n",
      " ('random', '', 758),\n",
      " ('random', '', 759),\n",
      " ('random', '', 760),\n",
      " ('random', '', 761),\n",
      " ('random', '', 762),\n",
      " ('random', '', 763),\n",
      " ('random', '', 764),\n",
      " ('random', '', 765),\n",
      " ('random', '', 766),\n",
      " ('random', '', 767),\n",
      " ('random', '', 768),\n",
      " ('random', '', 769),\n",
      " ('random', '', 770),\n",
      " ('random', '', 771),\n",
      " ('random', '', 772),\n",
      " ('random', '', 773),\n",
      " ('random', '', 774),\n",
      " ('random', '', 775),\n",
      " ('random', '', 776),\n",
      " ('random', '', 777),\n",
      " ('random', '', 778),\n",
      " ('random', '', 779),\n",
      " ('random', '', 780),\n",
      " ('random', '', 781),\n",
      " ('random', '', 782),\n",
      " ('random', '', 783),\n",
      " ('random', '', 784),\n",
      " ('random', '', 785),\n",
      " ('random', '', 786),\n",
      " ('random', '', 787),\n",
      " ('random', '', 788),\n",
      " ('random', '', 789),\n",
      " ('random', '', 790),\n",
      " ('random', '', 791),\n",
      " ('random', '', 792),\n",
      " ('random', '', 793),\n",
      " ('random', '', 794),\n",
      " ('random', '', 795),\n",
      " ('random', '', 796),\n",
      " ('random', '', 797),\n",
      " ('random', '', 798),\n",
      " ('random', '', 799),\n",
      " ('random', '', 800),\n",
      " ('random', '', 801),\n",
      " ('random', '', 802),\n",
      " ('random', '', 803),\n",
      " ('random', '', 804),\n",
      " ('random', '', 805),\n",
      " ('random', '', 806),\n",
      " ('random', '', 807),\n",
      " ('random', '', 808),\n",
      " ('random', '', 809),\n",
      " ('random', '', 810),\n",
      " ('random', '', 811),\n",
      " ('random', '', 812),\n",
      " ('random', '', 813),\n",
      " ('random', '', 814),\n",
      " ('random', '', 815),\n",
      " ('random', '', 816),\n",
      " ('random', '', 817),\n",
      " ('random', '', 818),\n",
      " ('random', '', 819),\n",
      " ('random', '', 820),\n",
      " ('random', '', 821),\n",
      " ('random', '', 822),\n",
      " ('random', '', 823),\n",
      " ('random', '', 824),\n",
      " ('random', '', 825),\n",
      " ('random', '', 826),\n",
      " ('random', '', 827),\n",
      " ('random', '', 828),\n",
      " ('random', '', 829),\n",
      " ('random', '', 830),\n",
      " ('random', '', 831),\n",
      " ('random', '', 832),\n",
      " ('random', '', 833),\n",
      " ('random', '', 834),\n",
      " ('random', '', 835),\n",
      " ('random', '', 836),\n",
      " ('random', '', 837),\n",
      " ('random', '', 838),\n",
      " ('random', '', 839),\n",
      " ('random', '', 840),\n",
      " ('random', '', 841),\n",
      " ('random', '', 842),\n",
      " ('random', '', 843),\n",
      " ('random', '', 844),\n",
      " ('random', '', 845),\n",
      " ('random', '', 846),\n",
      " ('random', '', 847),\n",
      " ('random', '', 848),\n",
      " ('random', '', 849),\n",
      " ('random', '', 850),\n",
      " ('random', '', 851),\n",
      " ('random', '', 852),\n",
      " ('random', '', 853),\n",
      " ('random', '', 854),\n",
      " ('random', '', 855),\n",
      " ('random', '', 856),\n",
      " ('random', '', 857),\n",
      " ('random', '', 858),\n",
      " ('random', '', 859),\n",
      " ('random', '', 860),\n",
      " ('random', '', 861),\n",
      " ('random', '', 862),\n",
      " ('random', '', 863),\n",
      " ('random', '', 864),\n",
      " ('random', '', 865),\n",
      " ('random', '', 866),\n",
      " ('random', '', 867),\n",
      " ('random', '', 868),\n",
      " ('random', '', 869),\n",
      " ('random', '', 870),\n",
      " ('random', '', 871),\n",
      " ('random', '', 872),\n",
      " ('random', '', 873),\n",
      " ('random', '', 874),\n",
      " ('random', '', 875),\n",
      " ('random', '', 876),\n",
      " ('random', '', 877),\n",
      " ('random', '', 878),\n",
      " ('random', '', 879),\n",
      " ('random', '', 880),\n",
      " ('random', '', 881),\n",
      " ('random', '', 882),\n",
      " ('random', '', 883),\n",
      " ('random', '', 884),\n",
      " ('random', '', 885),\n",
      " ('random', '', 886),\n",
      " ('random', '', 887),\n",
      " ('random', '', 888),\n",
      " ('random', '', 889),\n",
      " ('random', '', 890),\n",
      " ('random', '', 891),\n",
      " ('random', '', 892),\n",
      " ('random', '', 893),\n",
      " ('random', '', 894),\n",
      " ('random', '', 895),\n",
      " ('random', '', 896),\n",
      " ('random', '', 897),\n",
      " ('random', '', 898),\n",
      " ('random', '', 899),\n",
      " ('random', '', 900),\n",
      " ('random', '', 901),\n",
      " ('random', '', 902),\n",
      " ('random', '', 903),\n",
      " ('random', '', 904),\n",
      " ('random', '', 905),\n",
      " ('random', '', 906),\n",
      " ('random', '', 907),\n",
      " ('random', '', 908),\n",
      " ('random', '', 909),\n",
      " ('random', '', 910),\n",
      " ('random', '', 911),\n",
      " ('random', '', 912),\n",
      " ('random', '', 913),\n",
      " ('random', '', 914),\n",
      " ('random', '', 915),\n",
      " ('random', '', 916),\n",
      " ('random', '', 917),\n",
      " ('random', '', 918),\n",
      " ('random', '', 919),\n",
      " ('random', '', 920),\n",
      " ('random', '', 921),\n",
      " ('random', '', 922),\n",
      " ('random', '', 923),\n",
      " ('random', '', 924),\n",
      " ('random', '', 925),\n",
      " ('random', '', 926),\n",
      " ('random', '', 927),\n",
      " ('random', '', 928),\n",
      " ('random', '', 929),\n",
      " ('random', '', 930),\n",
      " ('random', '', 931),\n",
      " ('random', '', 932),\n",
      " ('random', '', 933),\n",
      " ('random', '', 934),\n",
      " ('random', '', 935),\n",
      " ('random', '', 936),\n",
      " ('random', '', 937),\n",
      " ('random', '', 938),\n",
      " ('random', '', 939),\n",
      " ('random', '', 940),\n",
      " ('random', '', 941),\n",
      " ('random', '', 942),\n",
      " ('random', '', 943),\n",
      " ('random', '', 944),\n",
      " ('random', '', 945),\n",
      " ('random', '', 946),\n",
      " ('random', '', 947),\n",
      " ('random', '', 948),\n",
      " ('random', '', 949),\n",
      " ('random', '', 950),\n",
      " ('random', '', 951),\n",
      " ('random', '', 952),\n",
      " ('random', '', 953),\n",
      " ('random', '', 954),\n",
      " ('random', '', 955),\n",
      " ('random', '', 956),\n",
      " ('random', '', 957),\n",
      " ('random', '', 958),\n",
      " ('random', '', 959),\n",
      " ('random', '', 960),\n",
      " ('random', '', 961),\n",
      " ('random', '', 962),\n",
      " ('random', '', 963),\n",
      " ('random', '', 964),\n",
      " ('random', '', 965),\n",
      " ('random', '', 966),\n",
      " ('random', '', 967),\n",
      " ('random', '', 968),\n",
      " ('random', '', 969),\n",
      " ('random', '', 970),\n",
      " ('random', '', 971),\n",
      " ('random', '', 972),\n",
      " ('random', '', 973),\n",
      " ('random', '', 974),\n",
      " ('random', '', 975),\n",
      " ('random', '', 976),\n",
      " ('random', '', 977),\n",
      " ('random', '', 978),\n",
      " ('random', '', 979),\n",
      " ('random', '', 980),\n",
      " ('random', '', 981),\n",
      " ('random', '', 982),\n",
      " ('random', '', 983),\n",
      " ('random', '', 984),\n",
      " ('random', '', 985),\n",
      " ('random', '', 986),\n",
      " ('random', '', 987),\n",
      " ('random', '', 988),\n",
      " ('random', '', 989),\n",
      " ('random', '', 990),\n",
      " ('random', '', 991),\n",
      " ('random', '', 992),\n",
      " ('random', '', 993),\n",
      " ('random', '', 994),\n",
      " ('random', '', 995),\n",
      " ('random', '', 996),\n",
      " ('random', '', 997),\n",
      " ('random', '', 998),\n",
      " ('random', '', 999),\n",
      " ('random', '', 1000),\n",
      " ('transformer-xl', '', 0)}\n"
     ]
    }
   ],
   "source": [
    "# Each model--corpus--seed should have perplexity data.\n",
    "ids_from_results = results_df.set_index([\"model_name\", \"corpus\", \"seed\"]).sort_index().index\n",
    "ids_from_ppl = perplexity_df.sort_index().index\n",
    "diff = set(ids_from_results) - set(ids_from_ppl)\n",
    "if diff:\n",
    "    print(\"Missing perplexity results for:\")\n",
    "    pprint(diff)\n",
    "    #raise ValueError(\"Each model--corpus--seed must have perplexity data.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OK\n"
     ]
    }
   ],
   "source": [
    "# Every model--corpus--seed should have results for all test suite items.\n",
    "item_list = {model_key: set(results.suite)\n",
    "             for model_key, results in results_df.groupby([\"model_name\", \"corpus\", \"seed\"])}\n",
    "not_shared = set()\n",
    "for k1, k2 in itertools.combinations(item_list.keys(), 2):\n",
    "    l1, l2 = item_list[k1], item_list[k2]\n",
    "    if l1 != l2:\n",
    "        print(\"SyntaxGym test suite results for %s and %s don't match\" % (k1, k2))\n",
    "        print(\"\\tIn %s but not in %s:\\n\\t\\t%s\" % (k2, k1, l2 - l1))\n",
    "        print(\"\\tIn %s but not in %s:\\n\\t\\t%s\" % (k1, k2, l1 - l2))\n",
    "        print()\n",
    "        \n",
    "        not_shared |= l2 - l1\n",
    "        not_shared |= l1 - l2\n",
    "\n",
    "if len(not_shared) > 0:\n",
    "    to_drop = results_df[results_df.suite.isin(not_shared)]\n",
    "    print(\"Dropping these test suites (%i rows) for now. Yikes:\" % len(to_drop))\n",
    "    print(not_shared)\n",
    "    results_df = results_df[~results_df.suite.isin(not_shared)]\n",
    "else:\n",
    "    print(\"OK\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Second sanity check: same number of results per model--corpus--seed\n",
    "result_counts = results_df.groupby([\"model_name\", \"corpus\", \"seed\"]).item.count()\n",
    "if len(result_counts.unique()) > 1:\n",
    "    print(\"WARNING: Some model--corpus--seed combinations have more result rows in results_df than others.\")\n",
    "    print(result_counts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Second sanity check: same number of suite-level results per model--corpus--seed\n",
    "suite_result_counts = suites_df.groupby([\"model_name\", \"corpus\", \"seed\"]).suite.count()\n",
    "if len(suite_result_counts.unique()) > 1:\n",
    "    print(\"WARNING: Some model--corpus--seed combinations have more result rows in suites_df than others.\")\n",
    "    print(suite_result_counts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare for data rendering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "RENDER_FINAL = True\n",
    "figure_path = Path(\"../reports/camera_ready_figures\")\n",
    "figure_path.mkdir(exist_ok=True, parents=True)\n",
    "\n",
    "RENDER_CONTEXT = {\n",
    "    \"font_scale\": 3.5,\n",
    "    \"rc\": {\"lines.linewidth\": 2.5, \"hatch.linewidth\":3},\n",
    "    \"style\": \"ticks\",\n",
    "    \"font\": \"Liberation Sans\"\n",
    "}\n",
    "sns.set(**RENDER_CONTEXT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "BASELINE_LINESTYLE = {\n",
    "    \"color\": \"gray\",\n",
    "    \"linestyle\": \"--\",\n",
    "}\n",
    "CORPUS_MARKERS = {\n",
    "    \"BLLIP-LG\": \"s\",\n",
    "    \"BLLIP-MD\": \"v\",\n",
    "    \"BLLIP-SM\": \"P\",\n",
    "    \"BLLIP-XS\": \"X\",\n",
    "    \n",
    "    \"BLLIP-LG-BPE\": \"s\",\n",
    "    \"BLLIP-MD-BPE\": \"v\",\n",
    "\n",
    "    \"BLLIP-LG-GPTBPE\": \"s\",\n",
    "    \"BLLIP-MD-GPTBPE\": \"v\",\n",
    "    \"BLLIP-SM-GPTBPE\": \"P\",\n",
    "    \"BLLIP-XS-GPTBPE\": \"X\"\n",
    "}\n",
    "p = sns.color_palette()[:len(joined_data.model_name.unique())]\n",
    "MODEL_COLORS = {\n",
    "    \"GPT-2\": p[0],\n",
    "    \"LSTM\": p[1],\n",
    "    \"ON-LSTM\": p[2],\n",
    "    \"RNNG\": p[3],\n",
    "    \"n-gram\": p[4],\n",
    "    \"Random\": \"darkgrey\",\n",
    "         \n",
    "    format_pretrained(\"GPT-2\"): \"mediumturquoise\",\n",
    "    format_pretrained(\"GPT-2-XL\"): p[5],\n",
    "    format_pretrained(\"Transformer-XL\"): \"gold\",\n",
    "    format_pretrained(\"GRNN\"): p[6],\n",
    "    format_pretrained(\"JRNN\"):  \"deeppink\",\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def render_final(path):\n",
    "    sns.despine()\n",
    "    plt.tight_layout()\n",
    "    plt.savefig(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Standardize axis labels\n",
    "SG_ABSOLUTE_LABEL = \"SG score\"\n",
    "SG_DELTA_LABEL = \"SG score delta\"\n",
    "PERPLEXITY_LABEL = \"Test perplexity\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Establish consistent orderings of model names, corpus names, circuit names\n",
    "# for figure ordering / coloring. (NB these refer to prettified names)\n",
    "model_order = sorted(set(results_df.pretty_model_name))\n",
    "controlled_model_order = [\"LSTM\", \"ON-LSTM\", \"RNNG\", \"GPT-2\", \"n-gram\"] #sorted(set(results_df[results_df.model_name.isin(controlled_models)].pretty_model_name))\n",
    "corpus_order = [\"BLLIP-LG\", \"BLLIP-MD\", \"BLLIP-SM\", \"BLLIP-XS\",\n",
    "                \"BLLIP-LG-BPE\", \"BLLIP-LG-GPTBPE\", \n",
    "                \"BLLIP-MD-GPTBPE\", \"BLLIP-SM-GPTBPE\", \"BLLIP-XS-GPTBPE\"]\n",
    "corpus_size_order = [\"lg\", \"md\", \"sm\", \"xs\"]\n",
    "nobpe_corpus_order = [c for c in corpus_order if \"BPE\" not in c]\n",
    "circuit_order = sorted([c for c in results_df.circuit.dropna().unique()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reproducing paper figures"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 1 (Basic barplot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "f, ax = plt.subplots(figsize=(20, 10))\n",
    "\n",
    "# Exclude random baseline; will plot as horizontal line\n",
    "plot_df = suites_df[(suites_df.model_name != \"random\")]\n",
    "\n",
    "# Sort by decreasing average accuracy\n",
    "order = list(plot_df.groupby(\"pretty_model_name\").correct.mean().sort_values(ascending=False).index)\n",
    "\n",
    "sns.barplot(data=plot_df.reset_index(), x=\"pretty_model_name\", y=\"correct\", order=order, ax=ax, palette=MODEL_COLORS)\n",
    "\n",
    "# Plot random chance baseline\n",
    "ax.axhline(suites_df[suites_df.model_name == \"random\"].correct.mean(), **BASELINE_LINESTYLE)\n",
    "\n",
    "# Adjust labels and axes\n",
    "ax.set_xticklabels(ax.get_xticklabels(), rotation=340, horizontalalignment=\"left\")\n",
    "ax.set_ylim(0,1)\n",
    "plt.xlabel(\"Model\")\n",
    "plt.ylabel(SG_ABSOLUTE_LABEL, labelpad=36)\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    render_final(figure_path / \"overall.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Controlled evaluation of model type + dataset size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "controlled_suites_df = suites_df[suites_df.model_name.isin(controlled_models)]\n",
    "controlled_suites_df_mod = suites_df_mod[suites_df_mod.model_name.isin(controlled_models)]\n",
    "controlled_joined_data_circuits = joined_data_circuits[joined_data_circuits.model_name.isin(controlled_models)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, axes = plt.subplots(nrows=1, ncols=2, sharex=False, sharey=True, figsize=(40,12))\n",
    "for i, ax in enumerate(axes):\n",
    "    ax.axhline(0, c=\"gray\", linestyle=\"--\")\n",
    "    if i == 0:\n",
    "        kwargs = dict(data=controlled_suites_df.reset_index(), order=controlled_model_order, ax=ax,\n",
    "                      x=\"pretty_model_name\", y=\"correct_delta\", palette=MODEL_COLORS)\n",
    "        sns.barplot(**kwargs, units=\"corpus\")\n",
    "        sns.swarmplot(**kwargs, alpha=0.3, size=9)\n",
    "\n",
    "        ax.set_xlabel(\"Model\", labelpad=16)\n",
    "        ax.set_ylabel(SG_DELTA_LABEL)\n",
    "    elif i == 1:\n",
    "        # Estimate error intervals with a structured bootstrap: resampling units = model\n",
    "        kwargs = dict(data=controlled_suites_df.reset_index(), x=\"pretty_corpus\", y=\"correct_delta\", order=nobpe_corpus_order, ax=ax)\n",
    "        sns.barplot(**kwargs, color=\"Gray\", units=\"pretty_model_name\")\n",
    "        sns.swarmplot(**kwargs, hue=\"pretty_model_name\",  hue_order=controlled_model_order, palette=MODEL_COLORS, size=9, alpha=0.5)\n",
    "\n",
    "        handles, labels = ax.get_legend_handles_labels()\n",
    "        for h in handles:\n",
    "            h.set_sizes([300.0])\n",
    "\n",
    "        ax.set_xlabel(\"Corpus\", labelpad=16)\n",
    "        ax.set_ylabel(\"\")\n",
    "        ax.legend(handles, labels, loc=\"upper center\", ncol=5, columnspacing=0.3, handletextpad=0.01)\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    render_final(figure_path / \"controlled.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, axes = plt.subplots(nrows=1, ncols=2, sharex=True, sharey=True, figsize=(40,15))\n",
    "legend_params=dict(title=\"\", ncol=5, loc=\"upper center\", columnspacing=1, handlelength=1, handletextpad=0.3)\n",
    "\n",
    "for i, ax in enumerate(axes):\n",
    "    ax.axhline(0, **BASELINE_LINESTYLE)\n",
    "    if i == 0:\n",
    "        sns.barplot(data=controlled_joined_data_circuits, x=\"circuit\", y=\"correct_delta\",\n",
    "                    hue=\"pretty_model_name\", units=\"corpus\", hue_order=controlled_model_order,\n",
    "                    ax=ax, palette=MODEL_COLORS)\n",
    "        ax.set_ylabel(SG_DELTA_LABEL)\n",
    "    elif i == 1:\n",
    "        sns.barplot(data=controlled_joined_data_circuits, x=\"circuit\", y=\"correct_delta\",\n",
    "                    hue=\"pretty_corpus\", units=\"model_name\", hue_order=nobpe_corpus_order,\n",
    "                    ax=ax, palette=\"Greys_r\")\n",
    "        ax.set_ylabel(\"\")\n",
    "        \n",
    "    ax.set_xticklabels(ax.get_xticklabels(), rotation=15, ha=\"right\")\n",
    "    ax.set_xlabel(\"Circuit\")\n",
    "    ax.legend(**legend_params)\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    render_final(figure_path / \"controlled_circuit.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, ax = plt.subplots(figsize=(40,12))\n",
    "joined_data_circuits_norandom = joined_data_circuits[joined_data_circuits.pretty_model_name != \"Random\"]\n",
    "order = list(plot_df.groupby(\"pretty_model_name\").correct.mean().sort_values(ascending=False).index)\n",
    "sns.barplot(data=joined_data_circuits_norandom, x=\"circuit\", y=\"correct\",\n",
    "            hue=\"pretty_model_name\", units=\"corpus\", hue_order=order, ax=ax, palette=MODEL_COLORS)\n",
    "\n",
    "ax.set_xticklabels(ax.get_xticklabels(), rotation=15, ha=\"right\")\n",
    "ax.set_xlabel(\"Circuit\")\n",
    "ax.set_ylabel(SG_ABSOLUTE_LABEL)\n",
    "ax.legend(title=\"\", ncol=int(len(order)/2), loc=\"upper center\", columnspacing=1, handlelength=1, handletextpad=1, bbox_to_anchor=(0.5,1.3))\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    render_final(figure_path / \"allmodels_circuit.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 6 (Stability to modification)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Suites in modification analysis:\", controlled_suites_df_mod.suite.unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sort by decreasing average accuracy.\n",
    "order = list(plot_df.groupby(\"pretty_model_name\").correct.mean().sort_values(ascending=False).index)\n",
    "\n",
    "_, ax = plt.subplots(figsize=(20,12))\n",
    "sns.barplot(data=suites_df_mod, x=\"pretty_model_name\", y=\"correct\", hue=\"has_modifier\", order=order, ax=ax)\n",
    "\n",
    "# Colors.\n",
    "sorted_patches = sorted(ax.patches, key=lambda bar: bar.get_x())\n",
    "colors = [MODEL_COLORS[order[i]] for i in range(len(order))]\n",
    "for i, bar in enumerate(sorted_patches):\n",
    "    bar.set_facecolor(colors[int(i/2)])\n",
    "    if i % 2 != 0:\n",
    "        bar.set_alpha(0.4)\n",
    "\n",
    "# Set labels.\n",
    "ax.set_xlabel(\"Model\", labelpad=16)\n",
    "ax.set_ylabel(SG_ABSOLUTE_LABEL)\n",
    "ax.set_xticklabels(ax.get_xticklabels(), rotation=340, horizontalalignment=\"left\")\n",
    "\n",
    "# Custom legend.\n",
    "handles, _ = ax.get_legend_handles_labels()\n",
    "handles[0] = mpatches.Patch(facecolor=\"k\")\n",
    "handles[1] = mpatches.Patch(facecolor=\"k\", alpha=0.4)\n",
    "ax.legend(handles, [\"No modifier\", \"With modifier\"], loc=\"upper right\", title=\"\")\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    render_final(figure_path / \"stability-all-models.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 2 (SG score vs perplexity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1368x1440 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Set limits for broken x-axis to  determine proper scaling (ratio of widths).\n",
    "ax1max = 250\n",
    "ax2min, ax2max = 520, 540\n",
    "ax_ratio = ax1max / (ax2max - ax2min)\n",
    "f, (ax1,ax2) = plt.subplots(1,2,sharey=False,figsize=(19, 20),gridspec_kw={'width_ratios': [ax_ratio, 1]})\n",
    "\n",
    "sns.despine()\n",
    "palette = sns.cubehelix_palette(4, reverse=True)\n",
    "\n",
    "markers = {\n",
    "    \"GPT-2\": \"s\",\n",
    "    \"RNNG\" : \"X\",\n",
    "    \"ON-LSTM\" : \"v\",\n",
    "    \"LSTM\" : \"*\",\n",
    "    \"n-gram\" : \"d\"\n",
    "}\n",
    "for m in joined_data.pretty_model_name.unique():\n",
    "    if m not in markers:\n",
    "        markers[m] = \".\"\n",
    "\n",
    "for ax in [ax1,ax2]:\n",
    "    sns.scatterplot(data=joined_data, x=\"test_ppl\", y=\"correct\", hue=\"corpus_size\", hue_order=corpus_size_order,\n",
    "                    markers=markers, palette=palette, style_order=model_order,\n",
    "                    s=2300, style=\"pretty_model_name\", ax=ax, zorder=2, alpha=0.8)\n",
    "    ax.set_xlabel(\"\")\n",
    "    ax.tick_params(axis='x', which='major', pad=15)\n",
    "\n",
    "    # Add horizontal lines for models without ppl estimates.\n",
    "    no_ppl_data = joined_data[joined_data.test_ppl.isna()]\n",
    "    for model_name, rows in no_ppl_data.groupby(\"pretty_model_name\"):\n",
    "        y = rows.correct.mean()\n",
    "        ax.axhline(y, zorder=1, linewidth=3, **BASELINE_LINESTYLE) \n",
    "        if \"GRNN\" in model_name: # custom spacing tweaking\n",
    "            y_offset = -0.03\n",
    "        else:\n",
    "            y_offset = 0.006\n",
    "        ax2.text(540, y + y_offset, model_name, fontdict={\"size\": 38}, ha='right')\n",
    "    \n",
    "plt.subplots_adjust(wspace=0.2)\n",
    "ax1.get_legend().remove()\n",
    "ax1.set_ylabel(SG_ABSOLUTE_LABEL)\n",
    "ax2.set_ylabel(\"\")\n",
    "plt.xlabel(PERPLEXITY_LABEL, labelpad=10, position=(-6,0))\n",
    "\n",
    "# Add break in x-axis\n",
    "ax1.set_xlim(0,ax1max)\n",
    "ax2.set_xlim(ax2min,ax2max)\n",
    "# hide the spines between ax1 and ax2\n",
    "ax1.spines['right'].set_visible(False)\n",
    "ax2.spines['left'].set_visible(False)\n",
    "ax2.get_yaxis().set_ticks([])\n",
    "d = .015 # how big to make the diagonal lines in axes coordinates\n",
    "kwargs = dict(transform=ax1.transAxes, color='k', clip_on=False)\n",
    "ax1.plot((1-d,1+d), (-d,+d), **kwargs)\n",
    "kwargs.update(transform=ax2.transAxes)  # switch to the right subplot\n",
    "ax2.plot((-d*ax_ratio,+d*ax_ratio), (-d,+d), **kwargs)\n",
    "    \n",
    "# Change some legend labels.\n",
    "handles, labels = ax1.get_legend_handles_labels()\n",
    "legend_title_map = {\"pretty_model_name\": \"Model\",\n",
    "                    \"pretty_corpus\": \"Corpus\",\n",
    "                    \"corpus_size\": \"Corpus size\",\n",
    "                    \"corpus_bpe\": \"Tokenization\"}\n",
    "# Re-map some labels.\n",
    "# labels = [legend_title_map.get(l, l) for l in labels]\n",
    "drop_indices = [i for i,l in enumerate(labels) if l in legend_title_map.keys() or l in no_ppl_data.pretty_model_name.values]\n",
    "handles = [h for i,h in enumerate(handles) if i not in drop_indices]\n",
    "labels = [l for i,l in enumerate(labels) if i not in drop_indices]\n",
    "labels = [l if l not in joined_data.corpus_size.unique() else \"BLLIP-%s\" % l.upper() for l in labels]\n",
    "\n",
    "# Add empty handle for legend spacing.\n",
    "handles.insert(4, mpatches.Patch(facecolor=\"white\"))\n",
    "labels.insert(4, \"\")\n",
    "\n",
    "# Re-order labels.\n",
    "new_order = [\"BLLIP-LG\", \"LSTM\", \"BLLIP-MD\", \"ON-LSTM\", \"BLLIP-SM\", \"RNNG\", \"BLLIP-XS\", \"GPT-2\", \"\", \"n-gram\"]\n",
    "inds = [labels.index(l) for l in new_order]\n",
    "handles = [handles[i] for i in inds]\n",
    "labels = [labels[i] for i in inds]\n",
    "\n",
    "# Set model style markers in legend to outlines only.\n",
    "for i, (l, h) in enumerate(zip(labels, handles)):\n",
    "    if l != \"\":\n",
    "        h.set_sizes([500.0])\n",
    "        if l in joined_data.pretty_model_name.unique():\n",
    "             handles[i] = Line2D([0], [0], marker=markers[l], color='k', mew=3, lw=0,\n",
    "                          markerfacecolor='w', markersize=27)\n",
    "\n",
    "plt.legend(handles, labels, bbox_to_anchor=(-16.4,-0.18), ncol=5, loc=\"center left\", columnspacing=0.5, handletextpad=0.05)\n",
    "\n",
    "if RENDER_FINAL:\n",
    "    # Can't use render_final function because of some spine issues.\n",
    "    plt.savefig(figure_path / \"perplexity.pdf\", bbox_inches=\"tight\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.3"
  },
  "toc-autonumbering": false,
  "toc-showcode": false,
  "toc-showmarkdowntxt": false
 },
 "nbformat": 4,
 "nbformat_minor": 4
}