{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Graph tensor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from molgraph import chemistry\n",
    "from molgraph import layers\n",
    "from molgraph import GraphTensor #####\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Construct a `GraphTensor`\n",
    "\n",
    "Although a `GraphTensor` can be constructed directly from its constructor, here we construct a `GraphTensor` from a `MolecularGraphEncoder`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 11), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_feature=<tf.Tensor: shape=(76, 5), dtype=float32>)\n"
     ]
    }
   ],
   "source": [
    "atom_encoder = chemistry.Featurizer([\n",
    "    chemistry.features.Symbol({'C', 'N', 'O'}, oov_size=1),\n",
    "    chemistry.features.Hybridization({'SP', 'SP2', 'SP3'}, oov_size=1),\n",
    "    chemistry.features.HydrogenDonor(),\n",
    "    chemistry.features.HydrogenAcceptor(),\n",
    "    chemistry.features.Hetero()\n",
    "])\n",
    "\n",
    "bond_encoder = chemistry.Featurizer([\n",
    "    chemistry.features.BondType({'SINGLE', 'DOUBLE', 'TRIPLE', 'AROMATIC'}),\n",
    "    chemistry.features.Rotatable()\n",
    "])\n",
    "\n",
    "mol_encoder = chemistry.MolecularGraphEncoder(\n",
    "    atom_encoder, bond_encoder, positional_encoding_dim=None)\n",
    "\n",
    "smiles_list = [\n",
    "    'OCC1OC(C(C1O)O)n1cnc2c1ncnc2N',\n",
    "    'C(C(=O)O)N',\n",
    "    'C1=CC(=CC=C1CC(C(=O)O)N)O'\n",
    "]\n",
    "\n",
    "graph_tensor = mol_encoder(smiles_list)\n",
    "\n",
    "print(graph_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.separate()` &ndash; Separate subgraphs of `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.RaggedTensor: shape=(3, None, 11), dtype=float32, ragged_rank=1>,\n",
      "  edge_src=<tf.RaggedTensor: shape=(3, None), dtype=int32, ragged_rank=1>,\n",
      "  edge_dst=<tf.RaggedTensor: shape=(3, None), dtype=int32, ragged_rank=1>,\n",
      "  edge_feature=<tf.RaggedTensor: shape=(3, None, 5), dtype=float32, ragged_rank=1>)\n"
     ]
    }
   ],
   "source": [
    "graph_tensor = graph_tensor.separate()\n",
    "print(graph_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.merge()` &ndash; Merge subgraphs of `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 11), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_feature=<tf.Tensor: shape=(76, 5), dtype=float32>)\n"
     ]
    }
   ],
   "source": [
    "graph_tensor = graph_tensor.merge()\n",
    "print(graph_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.propagate()` &ndash; Propagate node informaton with the `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Node features before:\n",
      " tf.Tensor(\n",
      "[[0. 0. 0. 1. 0. 0. 0. 1. 1. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 0. 0. 1. 0. 0. 0. 1. 0. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 0. 0. 1. 0. 0. 0. 1. 1. 1. 1.]\n",
      " [0. 0. 0. 1. 0. 0. 0. 1. 1. 1. 1.]\n",
      " [0. 0. 1. 0. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 1. 0. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 1. 0. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 1. 0. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 1. 0. 0. 0. 1. 0. 1. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 0. 1. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 0. 0. 1. 0. 0. 1. 0. 1. 0. 1.]\n",
      " [0. 0. 1. 0. 0. 0. 0. 1. 1. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 0. 0. 1. 0. 0. 1. 0. 0. 1. 1.]\n",
      " [0. 0. 0. 1. 0. 0. 1. 0. 1. 0. 1.]\n",
      " [0. 0. 1. 0. 0. 0. 0. 1. 1. 1. 1.]\n",
      " [0. 0. 0. 1. 0. 0. 1. 0. 1. 1. 1.]], shape=(37, 11), dtype=float32)\n",
      "\n",
      "Node features after:\n",
      " tf.Tensor(\n",
      "[[0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 1. 0. 0. 0. 2. 1. 1. 1.]\n",
      " [0. 2. 0. 1. 0. 0. 0. 3. 0. 1. 1.]\n",
      " [0. 2. 0. 0. 0. 0. 0. 2. 0. 0. 0.]\n",
      " [0. 1. 1. 1. 0. 0. 1. 2. 0. 2. 2.]\n",
      " [0. 2. 0. 1. 0. 0. 0. 3. 1. 1. 1.]\n",
      " [0. 2. 0. 1. 0. 0. 0. 3. 1. 1. 1.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 3. 0. 0. 0. 0. 2. 1. 0. 0. 0.]\n",
      " [0. 0. 2. 0. 0. 0. 2. 0. 0. 2. 2.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 2. 1. 0. 0. 0. 3. 0. 0. 1. 1.]\n",
      " [0. 1. 2. 0. 0. 0. 3. 0. 0. 2. 2.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 0. 2. 0. 0. 0. 2. 0. 0. 2. 2.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 1. 2. 0. 0. 0. 3. 0. 1. 2. 2.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 1. 0. 0. 0. 1. 1. 1. 1. 1.]\n",
      " [0. 1. 0. 2. 0. 0. 2. 1. 1. 1. 2.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 2. 0. 1. 0. 0. 3. 0. 1. 1. 1.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 2. 0. 0. 0. 0. 2. 0. 0. 0. 0.]\n",
      " [0. 3. 0. 0. 0. 0. 2. 1. 0. 0. 0.]\n",
      " [0. 2. 0. 0. 0. 0. 1. 1. 0. 0. 0.]\n",
      " [0. 2. 1. 0. 0. 0. 1. 2. 1. 1. 1.]\n",
      " [0. 1. 0. 2. 0. 0. 2. 1. 1. 1. 2.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 0. 1. 0. 0. 0.]\n",
      " [0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.]], shape=(37, 11), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "print('Node features before:\\n', graph_tensor.node_feature, end='\\n\\n')\n",
    "graph_tensor = graph_tensor.propagate()\n",
    "print('Node features after:\\n', graph_tensor.node_feature)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.update()` &ndash; Update data of the `GraphTensor`\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_feature=<tf.Tensor: shape=(76, 5), dtype=float32>,\n",
      "  node_supplementary_data=<tf.Tensor: shape=(37, 1), dtype=float32>)\n"
     ]
    }
   ],
   "source": [
    "node_supplementary_data = tf.random.uniform(\n",
    "    shape=graph_tensor.node_feature.shape[:-1] + [1])\n",
    "\n",
    "node_feature_updated = tf.random.uniform(\n",
    "    shape=graph_tensor.node_feature.shape[:-1] + [128])\n",
    "\n",
    "# Both add new data and update existing data of the GraphTensor:\n",
    "graph_tensor = graph_tensor.update({\n",
    "    'node_supplementary_data': node_supplementary_data, \n",
    "    'node_feature': node_feature_updated\n",
    "})\n",
    "print(graph_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.remove()` &ndash; Remove data from `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>)\n"
     ]
    }
   ],
   "source": [
    "graph_tensor = graph_tensor.remove([\n",
    "    'node_supplementary_data', 'edge_feature'\n",
    "])\n",
    "print(graph_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `__getitem__` &ndash; Index lookup on the `GraphTensor`\n",
    "\n",
    "The `GraphTensor` can be indexed either by passing a `str` (to obtain a specific field of `GraphTensor`) or `int`, `list[int]` or `slice` (to extract specific subgraphs (molecules) from `GraphTensor`). (Alternatively, the `GraphTensor` can be passed to `tf.gather` to extract specific subgraphs.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Complete graph:\n",
      "\n",
      "------------------------------------------------------------\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>)\n",
      "\n",
      "------------------------------------------------------------\n",
      "Subgraph (2) and (3) of graph:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(2,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(18, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(34,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(34,), dtype=int32>)\n",
      "\n",
      "------------------------------------------------------------\n",
      "Subgraph (2) and (3) of graph:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(2,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(24, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(50,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(50,), dtype=int32>)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Complete graph:\\n\")\n",
    "print(\"---\" * 20)\n",
    "print(graph_tensor, end='\\n\\n')\n",
    "\n",
    "print(\"---\" * 20)\n",
    "print(\"Subgraph (2) and (3) of graph:\\n\")\n",
    "print(graph_tensor[[1, 2]], end='\\n\\n')\n",
    "\n",
    "print(\"---\" * 20)\n",
    "print(\"Subgraph (2) and (3) of graph:\\n\")\n",
    "print(graph_tensor[:2], end='\\n\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `__getattr__` &ndash; Attribute lookup on the `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Access `node_feature` field:\n",
      "\n",
      "------------------------------------------------------------\n",
      "tf.Tensor(\n",
      "[[0.30606592 0.01332998 0.28550065 ... 0.30522108 0.43709052 0.2496804 ]\n",
      " [0.47505558 0.6802629  0.12628877 ... 0.54731417 0.85908985 0.01080072]\n",
      " [0.32505012 0.16541815 0.9268564  ... 0.19977057 0.6975106  0.63107324]\n",
      " ...\n",
      " [0.06981373 0.0497787  0.7329197  ... 0.72168195 0.992267   0.4002931 ]\n",
      " [0.6254629  0.77454865 0.4750824  ... 0.21217322 0.10769343 0.71567035]\n",
      " [0.29524624 0.7836231  0.7198993  ... 0.94255567 0.926514   0.62505746]], shape=(37, 128), dtype=float32)\n",
      "\n",
      "------------------------------------------------------------\n",
      "Access `edge_src` field:\n",
      "\n",
      "tf.Tensor(\n",
      "[ 0  1  1  2  2  2  3  3  4  4  4  5  5  5  6  6  6  7  8  9  9  9 10 10\n",
      " 11 11 12 12 12 13 13 13 14 14 15 15 16 16 17 17 17 18 19 19 20 20 20 21\n",
      " 22 23 24 24 25 25 26 26 26 27 27 28 28 29 29 29 30 30 31 31 31 32 32 32\n",
      " 33 34 35 36], shape=(76,), dtype=int32)\n",
      "\n",
      "------------------------------------------------------------\n",
      "Access `graph_indicator` field:\n",
      "\n",
      "tf.Tensor([0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2], shape=(37,), dtype=int64)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Access `node_feature` field:\\n\")\n",
    "print(\"---\" * 20)\n",
    "print(graph_tensor.node_feature, end='\\n\\n')\n",
    "\n",
    "print(\"---\" * 20)\n",
    "print(\"Access `edge_src` field:\\n\")\n",
    "print(graph_tensor.edge_src, end='\\n\\n')\n",
    "\n",
    "print(\"---\" * 20)\n",
    "print(\"Access `graph_indicator` field:\\n\")\n",
    "print(graph_tensor.graph_indicator, end='\\n\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `tf.concat` &ndash; Concatenating multiple `GraphTensor` instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Concatenating two graphs in non-ragged states:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(6,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(74, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(152,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(152,), dtype=int32>)\n",
      "\n",
      "Inspect `graph_indicator`:\n",
      "\n",
      "tf.Tensor(\n",
      "[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2\n",
      " 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5], shape=(74,), dtype=int64)\n",
      "\n",
      "------------------------------------------------------------\n",
      "Concatenating two graphs in ragged states\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(6,), dtype=int64>,\n",
      "  node_feature=<tf.RaggedTensor: shape=(6, None, 128), dtype=float32, ragged_rank=1>,\n",
      "  edge_src=<tf.RaggedTensor: shape=(6, None), dtype=int32, ragged_rank=1>,\n",
      "  edge_dst=<tf.RaggedTensor: shape=(6, None), dtype=int32, ragged_rank=1>)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Concatenating two graphs in non-ragged states:\\n\")\n",
    "graph_tensor_concat = tf.concat([\n",
    "    graph_tensor, \n",
    "    graph_tensor], axis=0)\n",
    "print(graph_tensor_concat, end='\\n\\n')\n",
    "print(\"Inspect `graph_indicator`:\\n\")\n",
    "print(graph_tensor_concat.graph_indicator, end='\\n\\n')\n",
    "\n",
    "print('---' * 20)\n",
    "print(\"Concatenating two graphs in ragged states\")\n",
    "graph_tensor_concat = tf.concat([\n",
    "    graph_tensor.separate(), \n",
    "    graph_tensor.separate()], axis=0)\n",
    "print(graph_tensor_concat, end='\\n\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `tf.split` &ndash; Splits a `GraphTensor` into multiple `GraphTensor` instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(19, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(42,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(42,), dtype=int32>),\n",
       " GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(5, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(8,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(8,), dtype=int32>),\n",
       " GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(13, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(26,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(26,), dtype=int32>),\n",
       " GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(19, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(42,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(42,), dtype=int32>),\n",
       " GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(5, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(8,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(8,), dtype=int32>),\n",
       " GraphTensor(\n",
       "   sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
       "   node_feature=<tf.Tensor: shape=(13, 128), dtype=float32>,\n",
       "   edge_src=<tf.Tensor: shape=(26,), dtype=int32>,\n",
       "   edge_dst=<tf.Tensor: shape=(26,), dtype=int32>)]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.split(graph_tensor_concat.merge(), num_or_size_splits=6)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.spec` &ndash; The spec of the `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GraphTensor.Spec(sizes=TensorSpec(shape=(None,), dtype=tf.int64, name=None), node_feature=TensorSpec(shape=(None, 128), dtype=tf.float32, name=None), edge_src=TensorSpec(shape=(None,), dtype=tf.int32, name=None), edge_dst=TensorSpec(shape=(None,), dtype=tf.int32, name=None), edge_feature=None, edge_weight=None, node_position=None, auxiliary={})\n"
     ]
    }
   ],
   "source": [
    "print(graph_tensor.spec)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.shape` &ndash; Partial shape of the `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(partial) shape: (3, None, 128)\n"
     ]
    }
   ],
   "source": [
    "print('(partial) shape:', graph_tensor.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.dtype` &ndash; Partial dtype of the `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(partial) dtype: float32\n"
     ]
    }
   ],
   "source": [
    "print('(partial) dtype:', graph_tensor.dtype.name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `.rank` &ndash; Partial rank of the `GraphTensor` "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(partial) rank:  3\n"
     ]
    }
   ],
   "source": [
    "print('(partial) rank: ', graph_tensor.rank)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `tf.data.Dataset` &ndash; Creating a TF dataset from a `GraphTensor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset object:\n",
      " <_TensorSliceDataset element_spec=GraphTensor.Spec(sizes=TensorSpec(shape=(), dtype=tf.int64, name=None), node_feature=TensorSpec(shape=(None, 128), dtype=tf.float32, name=None), edge_src=TensorSpec(shape=(None,), dtype=tf.int32, name=None), edge_dst=TensorSpec(shape=(None,), dtype=tf.int32, name=None), edge_feature=None, edge_weight=None, node_position=None, auxiliary={})>\n",
      "\n",
      "------------------------------------------------------------\n",
      "\n",
      "batch 1:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(2,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(24, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(50,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(50,), dtype=int32>)\n",
      "\n",
      "------------------------------------------------------------\n",
      "\n",
      "batch 2:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(1,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(13, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(26,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(26,), dtype=int32>)\n",
      "\n",
      "------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "ds = tf.data.Dataset.from_tensor_slices(graph_tensor)\n",
    "print('Dataset object:\\n', ds)\n",
    "\n",
    "print('\\n' + '---' * 20)\n",
    "# Loop over dataset\n",
    "for i, x in enumerate(ds.batch(2).map(lambda x: x)):\n",
    "    print(f\"\\nbatch {i + 1}:\\n\")\n",
    "    print(x)\n",
    "    print('\\n' + '---' * 20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `layers` &ndash; Passing a `GraphTensor`  to a layer\n",
    "\n",
    "The `GraphTensor` can be passed to GNN layers either as a single disjoint graph or subgraphs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pass GraphTensor in non-ragged state:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.Tensor: shape=(37, 128), dtype=float32>,\n",
      "  edge_src=<tf.Tensor: shape=(76,), dtype=int32>,\n",
      "  edge_dst=<tf.Tensor: shape=(76,), dtype=int32>)\n",
      "\n",
      "------------------------------------------------------------\n",
      "\n",
      "Pass GraphTensor in ragged state:\n",
      "\n",
      "GraphTensor(\n",
      "  sizes=<tf.Tensor: shape=(3,), dtype=int64>,\n",
      "  node_feature=<tf.RaggedTensor: shape=(3, None, 128), dtype=float32, ragged_rank=1>,\n",
      "  edge_src=<tf.RaggedTensor: shape=(3, None), dtype=int32, ragged_rank=1>,\n",
      "  edge_dst=<tf.RaggedTensor: shape=(3, None), dtype=int32, ragged_rank=1>)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "gin_conv = layers.GINConv(128)\n",
    "\n",
    "print(\"Pass GraphTensor in non-ragged state:\\n\")\n",
    "print(gin_conv(graph_tensor), end='\\n\\n')\n",
    "print('---' * 20)\n",
    "print('\\nPass GraphTensor in ragged state:\\n')\n",
    "print(gin_conv(graph_tensor.separate()), end='\\n\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `Model` &ndash; Passing a `GraphTensor` to a model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using (graph_tensor, label) pair as input:\n",
      "\n",
      "Epoch 1/5\n",
      "2/2 [==============================] - 3s 8ms/step - loss: 0.3226\n",
      "Epoch 2/5\n",
      "2/2 [==============================] - 0s 7ms/step - loss: 8.0285\n",
      "Epoch 3/5\n",
      "2/2 [==============================] - 0s 8ms/step - loss: 4.7673\n",
      "Epoch 4/5\n",
      "2/2 [==============================] - 0s 6ms/step - loss: 1.8421\n",
      "Epoch 5/5\n",
      "2/2 [==============================] - 0s 6ms/step - loss: 0.1327\n",
      "\n",
      "------------------------------\n",
      "\n",
      "Using tf.data.Dataset as input:\n",
      "\n",
      "Epoch 1/5\n",
      "2/2 [==============================] - 0s 7ms/step - loss: 0.8891\n",
      "Epoch 2/5\n",
      "2/2 [==============================] - 0s 9ms/step - loss: 1.2614\n",
      "Epoch 3/5\n",
      "2/2 [==============================] - 0s 6ms/step - loss: 1.0532\n",
      "Epoch 4/5\n",
      "2/2 [==============================] - 0s 6ms/step - loss: 0.6833\n",
      "Epoch 5/5\n",
      "2/2 [==============================] - 0s 6ms/step - loss: 0.4317\n"
     ]
    }
   ],
   "source": [
    "model = tf.keras.Sequential([\n",
    "    layers.GCNConv(),\n",
    "    layers.GCNConv(),\n",
    "    layers.Readout(),\n",
    "    keras.layers.Dense(1)\n",
    "])\n",
    "\n",
    "y_dummy = tf.constant([[1.], [2.], [3.]])\n",
    "\n",
    "\n",
    "model.compile('adam', 'huber')\n",
    "print(\"Using (graph_tensor, label) pair as input:\\n\")\n",
    "model.fit(graph_tensor, y_dummy, batch_size=2, epochs=5)\n",
    "\n",
    "print('\\n------------------------------\\n')\n",
    "print(\"Using tf.data.Dataset as input:\\n\")\n",
    "dataset = tf.data.Dataset.from_tensor_slices((graph_tensor, y_dummy))\n",
    "model.fit(dataset.batch(2), epochs=5);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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": "3.10.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}