Datasets:

axd353
/

NoRA-1.1

+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 48,
+   "id": "43220e88",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from datasets import load_dataset\n",
+    "from huggingface_hub import HfApi\n",
+    "from huggingface_hub.utils import HfHubHTTPError\n",
+    "import pandas as pd\n",
+    "import json,ast\n",
+    "from itertools import islice\n",
+    "import pprint\n",
+    "pd.set_option('display.max_rows', None)\n",
+    "pd.set_option('display.max_columns', None)\n",
+    "# Don’t truncate column contents\n",
+    "pd.set_option('display.max_colwidth', None)\n",
+    "# Don’t wrap wide DataFrames\n",
+    "pd.set_option('display.expand_frame_repr', False)\n",
+    "pp = pprint.PrettyPrinter(width=200, compact=False)\n",
+    "pp = pprint.PrettyPrinter(width=200, compact=False)\n",
+    "#pip install \"datasets>=2.20\" \"huggingface_hub>=0.24\" \"pandas>=2.0\" jupyter\n",
+    "#paper When No Paths Lead to Rome: Benchmarking Systematic Neural Relational Reasoning. 39th Conference on Neural Information Processing Systems (NeurIPS 2025)."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 49,
+   "id": "32ac5a59",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'pandas.core.frame.DataFrame'>\n",
+      "Index(['Unnamed: 0', 'edge_types', 'story_edges', 'query_edge', 'query_label',\n",
+      "       'query_relation', 'derivation_chain', 'BL', 'OPEC', 'ReasoningDepth',\n",
+      "       'derivations_for_other_implied_relationships',\n",
+      "       'metric_for_other_implied_relationships', 'source_file',\n",
+      "       'original_row_index', 'alternate_labels_true'],\n",
+      "      dtype='object')\n",
+      "(10589, 15)\n"
+     ]
+    }
+   ],
+   "source": [
+    "##importing training data\n",
+    "repo_name =  f'NoRA-1.1'\n",
+    "ds_train = load_dataset(f\"axd353/{repo_name}\", split=\"train\")\n",
+    "df_train = ds_train.to_pandas()\n",
+    "print(type(df_train))\n",
+    "print(df_train.columns)\n",
+    "print(df_train.shape)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 52,
+   "id": "2135e566",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "##helper methods to explain and visualize\n",
+    "def dataframe_to_string_with_gaps(df: pd.DataFrame, output_file: str, column_list=None):\n",
+    "    \"\"\"\n",
+    "    Takes a Pandas DataFrame and writes each row to a text file as a string,\n",
+    "    with three line gaps between rows and one line gap between columns within a row.\n",
+    "    Does not truncate any content.\n",
+    "\n",
+    "    Args:\n",
+    "        df (pd.DataFrame): The input DataFrame.\n",
+    "        output_file (str): The name of the output text file.\n",
+    "    \"\"\"\n",
+    "    if column_list != None:\n",
+    "        df =df[column_list]\n",
+    "    with open(output_file, 'w', encoding='utf-8') as f:\n",
+    "        for index, row in df.iterrows():\n",
+    "            row_str = \"\"\n",
+    "            for col_name, value in row.items():\n",
+    "                row_str += f'{col_name}: '+str(value)\n",
+    "                row_str += \"\\n\"  # One line gap between columns\n",
+    "            f.write(row_str)\n",
+    "            f.write(\"\\n\\n\\n\")  # Three line gaps between rows\n",
+    "\n",
+    "def print_zipped_edges(row_df: pd.DataFrame, M: int = 6, sep: str = \"    \", blank_lines: int = 1) -> None:\n",
+    "    \"\"\"\n",
+    "    Print zipped (edge_types, story_edges) from a single row in chunks of M.\n",
+    "\n",
+    "    Parameters\n",
+    "    ----------\n",
+    "    row_df : pd.DataFrame or pd.Series\n",
+    "        Single-row DataFrame or Series with 'edge_types' and 'story_edges' as lists.\n",
+    "    M : int, default=6\n",
+    "        Tuples per line.\n",
+    "    sep : str, default=\"   \"\n",
+    "        Separator between tuples on the same line.\n",
+    "    blank_lines : int, default=1\n",
+    "        Number of blank lines inserted between printed lines.\n",
+    "    \"\"\"\n",
+    "    r = row_df.iloc[0] if isinstance(row_df, pd.DataFrame) else row_df\n",
+    "    et = r['edge_types']\n",
+    "    se = r['story_edges']\n",
+    "\n",
+    "    zipped = list(zip(et, se))\n",
+    "\n",
+    "    for start in range(0, len(zipped), M):\n",
+    "        chunk = zipped[start:start+M]\n",
+    "        # build one line without the list brackets, with custom separation between tuples\n",
+    "        line = sep.join(str(t) for t in chunk)\n",
+    "        print(line)\n",
+    "        if blank_lines > 0:\n",
+    "            print(\"\\n\" * blank_lines, end=\"\")\n",
+    "\n",
+    "\n",
+    "def _to_list(x):\n",
+    "    # minimal + robust: try JSON, then Python literal, then comma-split\n",
+    "    if isinstance(x, (list, tuple)): \n",
+    "        return list(x)\n",
+    "    if pd.isna(x):\n",
+    "        return []\n",
+    "    if isinstance(x, str):\n",
+    "        s = x.strip()\n",
+    "        for parser in (json.loads, ast.literal_eval):\n",
+    "            try:\n",
+    "                v = parser(s)\n",
+    "                return list(v) if isinstance(v, (list, tuple)) else [v]\n",
+    "            except Exception:\n",
+    "                pass\n",
+    "        # fallback: comma-separated\n",
+    "        return [p.strip() for p in s.split(\",\") if p.strip()]\n",
+    "    return [x]\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 53,
+   "id": "f608ac19",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "##some columns have bene turned into string getting them back to list\n",
+    "for col in ['edge_types', 'story_edges', 'query_label']:\n",
+    "    df_train[col] = df_train[col].apply(_to_list)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 57,
+   "id": "05650457",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "('is_person', (0, 0))    ('is_male', (0, 0))    ('is_person', (1, 1))    ('is_female', (1, 1))    ('is_person', (2, 2))    ('is_male', (2, 2))    ('is_place', (3, 3))    ('is_person', (4, 4))\n",
+      "\n",
+      "('is_female', (4, 4))    ('is_person', (5, 5))    ('is_female', (5, 5))    ('is_person', (6, 6))    ('is_female', (6, 6))    ('is_place', (7, 7))    ('is_place', (8, 8))    ('is_person', (9, 9))\n",
+      "\n",
+      "('is_person', (10, 10))    ('is_male', (10, 10))    ('is_person', (11, 11))    ('is_male', (11, 11))    ('is_person', (12, 12))    ('is_male', (12, 12))    ('is_person', (13, 13))    ('is_male', (13, 13))\n",
+      "\n",
+      "('is_person', (14, 14))    ('is_male', (14, 14))    ('is_person', (15, 15))    ('is_male', (15, 15))    ('is_place', (16, 16))    ('is_person', (17, 17))    ('is_male', (17, 17))    ('is_place', (18, 18))\n",
+      "\n",
+      "('is_person', (19, 19))    ('is_female', (19, 19))    ('is_underage', (19, 19))    ('is_person', (20, 20))    ('is_female', (20, 20))    ('is_place', (21, 21))    ('is_person', (22, 22))    ('is_male', (22, 22))\n",
+      "\n",
+      "('is_place', (23, 23))    ('no_sisters', (0, 0))    ('no_brothers', (1, 1))    ('no_sons', (15, 15))    ('living_in', (6, 3))    ('father_of', (11, 19))    ('living_in_same_place', (1, 15))    ('grandson_of', (22, 1))\n",
+      "\n",
+      "('living_in', (2, 23))    ('living_in', (20, 16))    ('school_mates_with', (17, 0))    ('living_in_same_place', (17, 11))    ('living_in', (19, 21))    ('nephew_of', (14, 6))    ('living_in', (1, 16))    ('grandparent_of', (13, 9))\n",
+      "\n",
+      "('school_mates_with', (20, 12))    ('living_in_same_place', (0, 5))    ('grandparent_of', (15, 5))    ('sibling_in_law_of', (19, 17))    ('grandparent_of', (14, 20))    ('living_in_same_place', (5, 13))    ('paternal_aunt_or_uncle_of', (12, 1))    ('paternal_grandparent_of', (22, 17))\n",
+      "\n",
+      "('living_in', (4, 8))    ('living_in', (14, 16))    ('aunt_or_uncle_of', (2, 0))    ('sister_in_law_of', (1, 11))    ('living_in', (9, 8))    ('parent_of', (4, 11))    ('aunt_of', (9, 15))    ('school_mates_with', (17, 10))\n",
+      "\n",
+      "('paternal_aunt_of', (4, 9))    ('brother_of', (10, 20))    ('living_in', (10, 21))    ('grandparent_of', (22, 20))    ('wife_of', (5, 6))    ('maternal_aunt_of', (19, 14))    ('uncle_of', (11, 22))    ('grandparent_of', (4, 12))\n",
+      "\n",
+      "('maternal_grandparent_of', (13, 1))    ('sibling_of', (15, 19))    ('living_in', (0, 21))    ('living_in', (22, 18))    ('maternal_grandparent_of', (13, 2))    ('living_in', (17, 21))    ('maternal_aunt_or_uncle_of', (4, 6))    ('aunt_or_uncle_of', (11, 5))\n",
+      "\n",
+      "('nibling_of', (5, 0))    ('father_in_law_of', (13, 0))    ('paternal_grandparent_of', (14, 10))\n",
+      "\n",
+      "5597    (15, 5)\n",
+      "Name: query_edge, dtype: object\n",
+      "5597    [grandfather_of, grandparent_of, maternal_grandparent_of, maternal_grandfather_of]\n",
+      "Name: query_label, dtype: object\n"
+     ]
+    }
+   ],
+   "source": [
+    "##description of task \n",
+    "row =  df_train[(df_train['source_file']=='result_59519031_8_seed244.csv') & (df_train['original_row_index']==21116)]\n",
+    "print_zipped_edges(row, M=8)\n",
+    "print(row['query_edge'])\n",
+    "print(row['query_label'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 58,
+   "id": "b6d92ae3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "explore_df2 = df_train[(df_train['source_file'] == 'result_59519031_8_seed244.csv') & (df_train['original_row_index']==21116)]\n",
+    "dataframe_to_string_with_gaps(explore_df2, '/home/anirban/CameraReadyVersionForNeurIPS/MakingVideos/stories2.txt')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "578b1225",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "##visualize story\n",
+    "row =  df_train[(df_train['source_file']=='result_59526051_6_seed296.csv') & (df_train['original_row_index']==483)]\n",
+    "print_zipped_edges(row, M=8)\n",
+    "print(row['query_edge'])\n",
+    "print(row['query_label'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "aff4091e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "##visualizing the derivation of labels in a story \n",
+    "explore_df = df_train[(df_train['source_file']=='result_59526051_6_seed296.csv') & (df_train['original_row_index']==483)]\n",
+    "dataframe_to_string_with_gaps(explore_df, '/home/anirban/CameraReadyVersionForNeurIPS/MakingVideos/stories1.txt')"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Edge Transformer Env",
+   "language": "python",
+   "name": "edge_transformer"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}