Add more notebooks assessing training effect

scripts/evaluate-generated-mofs/2_effect-of-training.ipynb

@@ -0,0 +1,232 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "5211ca75-c0ed-4e81-ae37-2a21533656f1",
+   "metadata": {},
+   "source": [
+    "# Evaluate the Effect of Training \n",
+    "We can assess whether retraining Difflinker leads to improved performance in two ways:\n",
+    "1. Evaluate how much the success rate improves with re-training\n",
+    "2. The difference between the total number of stable MOFs found w/ and w/o a closed loop"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "e0c564c2-ec6d-4d55-bc8e-944a80d35598",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from itertools import chain\n",
+    "from scipy.interpolate import interp1d\n",
+    "from pathlib import Path\n",
+    "import pandas as pd"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c56bd56d-a264-4e11-a7ec-e3bb62268e42",
+   "metadata": {},
+   "source": [
+    "## Route 1: Measure Success Rate by Model Generation"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c8bb1426-af24-4159-b1c4-55e9c203bfde",
+   "metadata": {},
+   "source": [
+    "## Round 2: Assess workflow outcomes w/o retraining\n",
+    "Show that it gets better"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "10478466-dfd8-41c0-bf53-20dcb394fb91",
+   "metadata": {},
+   "source": [
+    "### Get the \"Stable Found\" at 90 minutes\n",
+    "Loop over all runs and store: scale, if retrained or not, and the number of stable found after 90 minutes. \n",
+    "The 450-node run switches how it trained DiffLinker at around 90 minutes, and we don't want to study that effect yet."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 33,
+   "id": "aa46b598-7a7f-4f13-8a7c-c991ef4e2013",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "hours = 1.5"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 34,
+   "id": "98400659-f017-4b68-8ac7-1e9643c10c65",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "success_data = []\n",
+    "for path in chain(Path('summaries').glob('*-nodes.csv.gz'), Path('summaries').glob('*-nodes_repeat-*.csv.gz'), Path('summaries').glob('*no-retrain*.csv.gz')):\n",
+    "    # Get metadata\n",
+    "    count = int(path.name.split(\"-\")[0])\n",
+    "    retrain = 'no-retrain' not in path.name\n",
+    "\n",
+    "    # Pull the success rate\n",
+    "    mofs = pd.read_csv(path)\n",
+    "    num_found = interp1d(mofs['walltime'], mofs['cumulative_found'], kind='previous')(hours * 3600).item()\n",
+    "\n",
+    "    success_data.append({\n",
+    "        'nodes': count,\n",
+    "        'retrain': retrain,\n",
+    "        'found': num_found,\n",
+    "        'found_node-hr': num_found / (count * hours)\n",
+    "    })\n",
+    "success_data = pd.DataFrame(success_data)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 35,
+   "id": "ac0c533f-f071-4e74-95d1-77fb0a9e17d9",
+   "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></th>\n",
+       "      <th>found</th>\n",
+       "      <th>found_node-hr</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>nodes</th>\n",
+       "      <th>retrain</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th rowspan=\"2\" valign=\"top\">32</th>\n",
+       "      <th>False</th>\n",
+       "      <td>133.0</td>\n",
+       "      <td>2.770833</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>True</th>\n",
+       "      <td>313.0</td>\n",
+       "      <td>6.520833</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th rowspan=\"2\" valign=\"top\">64</th>\n",
+       "      <th>False</th>\n",
+       "      <td>426.5</td>\n",
+       "      <td>4.442708</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>True</th>\n",
+       "      <td>641.0</td>\n",
+       "      <td>6.677083</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>128</th>\n",
+       "      <th>True</th>\n",
+       "      <td>1622.0</td>\n",
+       "      <td>8.447917</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>256</th>\n",
+       "      <th>True</th>\n",
+       "      <td>3633.0</td>\n",
+       "      <td>9.460938</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>450</th>\n",
+       "      <th>True</th>\n",
+       "      <td>6554.0</td>\n",
+       "      <td>9.709630</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "                found  found_node-hr\n",
+       "nodes retrain                       \n",
+       "32    False     133.0       2.770833\n",
+       "      True      313.0       6.520833\n",
+       "64    False     426.5       4.442708\n",
+       "      True      641.0       6.677083\n",
+       "128   True     1622.0       8.447917\n",
+       "256   True     3633.0       9.460938\n",
+       "450   True     6554.0       9.709630"
+      ]
+     },
+     "execution_count": 35,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "success_data.groupby(['nodes', 'retrain']).mean()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "a973a7cb-e161-429e-9707-4f6b6a608bd2",
+   "metadata": {},
+   "source": [
+    "TBD: Make a plot"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "269787d5-5319-47f7-83ce-1a807bf14583",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "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": "3.10.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}