{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HDObRj0KtYYi"
   },
   "source": [
    "# Batch support\n",
    "\n",
    "This notebook explains how to use the batch support in Haiku Geometric.\n",
    "\n",
    "\n",
    "[![Open in Colab](https://img.shields.io/static/v1.svg?logo=google-colab&label=Quickstart&message=Open%20In%20Colab&color=blue)](https://colab.research.google.com/github/alexOarga/haiku-geometric/blob/main/docs/source/notebooks/batch_support.ipynb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install git+https://github.com/alexOarga/haiku-geometric.git"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_IyJz0r0tugA"
   },
   "source": [
    "## Batching graphs\n",
    "\n",
    "The `haiku_geometric.utils.batch` function can be used to batch a list of\n",
    "`haiku_geometric.data.DataGraphTuple` objects into a single `haiku_geometric.data.DataGraphTuple` object.\n",
    "\n",
    "The `batch` function returns:\n",
    "  - A single `haiku_geometric.data.DataGraphTuple` with the batched graphs.\n",
    "  - A `jax.numpy.Array` with indices indicating to which graph each node belongs to."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "krfCUf9HtoKt",
    "outputId": "46f3833d-9d0c-425b-e493-548d001cb198"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:jax._src.xla_bridge:No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
     ]
    }
   ],
   "source": [
    "import jax.numpy as jnp\n",
    "from haiku_geometric.utils import batch\n",
    "from haiku_geometric.datasets.base import DataGraphTuple\n",
    "\n",
    "graph1 = DataGraphTuple(\n",
    "    nodes=jnp.array([0.0, 0.1, 0.2]),\n",
    "    senders=jnp.array([0, 1, 2]),\n",
    "    receivers=jnp.array([2, 2, 0]),\n",
    "    edges=None,\n",
    "    n_node=jnp.array([3]),\n",
    "    n_edge=jnp.array([3]),\n",
    "    globals=None,\n",
    "    position=None,\n",
    "    y=jnp.array([0, 0, 0]),\n",
    "    train_mask=None,\n",
    ")\n",
    "\n",
    "graph2 = DataGraphTuple(\n",
    "    nodes=jnp.array([0.0, 0.0]),\n",
    "    senders=jnp.array([0, 1]),\n",
    "    receivers=jnp.array([1, 0]),\n",
    "    edges=None,\n",
    "    n_node=jnp.array([2]),\n",
    "    n_edge=jnp.array([2]),\n",
    "    globals=None,\n",
    "    position=None,\n",
    "    y=jnp.array([0, 0]),\n",
    "    train_mask=None,\n",
    ")\n",
    "\n",
    "batched_graph, batch_index = batch([graph1, graph2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bm3sA4o1uSn3"
   },
   "source": [
    "## Unbatching graphs\n",
    "\n",
    "To unbatch a `DataGraphTuple` object created with `batch` function,\n",
    "we can use the `haiku_geometric.utils.unbatch` function. This function takes a `DataGraphTuple`\n",
    "object and returns a list of `haiku_geometric.data.DataGraphTuple` objects."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "xyDiWJv5ugsr"
   },
   "outputs": [],
   "source": [
    "from haiku_geometric.utils import unbatch\n",
    "\n",
    "unbatched_graphs = unbatch(batched_graph)\n",
    "graph1 = unbatched_graphs[0]\n",
    "graph2 = unbatched_graphs[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UtgF9JP1ZZ1-"
   },
   "source": [
    "## Dynamic batching\n",
    "\n",
    "Unfortunately, Haiku Geometric does not currently support dynamic batching. If you are working with `jraph`,\n",
    "you can create a `jraph.GraphsTuple` object and use the available function `jraph.dynamically_batch`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "-HqgVgEvZawQ"
   },
   "outputs": [],
   "source": [
    "import jax.numpy as jnp\n",
    "import jraph\n",
    "\n",
    "graph1 = jraph.GraphsTuple(\n",
    "    nodes=jnp.array([0.0, 0.1, 0.2]),\n",
    "    senders=jnp.array([0, 1, 2]),\n",
    "    receivers=jnp.array([2, 2, 0]),\n",
    "    edges=None,\n",
    "    n_node=jnp.array([3]),\n",
    "    n_edge=jnp.array([3]),\n",
    "    globals=None,\n",
    ")\n",
    "\n",
    "graph2 = jraph.GraphsTuple(\n",
    "    nodes=jnp.array([0.0, 0.0]),\n",
    "    senders=jnp.array([0, 1]),\n",
    "    receivers=jnp.array([1, 0]),\n",
    "    edges=None,\n",
    "    n_node=jnp.array([2]),\n",
    "    n_edge=jnp.array([2]),\n",
    "    globals=None,\n",
    ")\n",
    "\n",
    "MAXIMUM_NUM_NODES = 2\n",
    "MAXIMUM_NUM_EDGES = 3\n",
    "MAXIMUM_NUM_GRAPHS = 2\n",
    "\n",
    "batched_generator = jraph.dynamically_batch([graph1, graph2],\n",
    "                                        MAXIMUM_NUM_NODES, # max number of nodes in a batch\n",
    "                                        MAXIMUM_NUM_EDGES, # max number of edges in a batch\n",
    "                                        MAXIMUM_NUM_GRAPHS)  # max number of graphs in a batch"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "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": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}