notebooks/tutorial_jax_jit.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "d9e621b0-f26e-461d-abc9-2da351a73a26",
   "metadata": {},
   "source": [
    "# Using JAX as Backend\n",
    "\n",
    "Showcase how to use jitable jax models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "27d6f797-5e9d-4dd3-b74d-ba09f666baea",
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras as ks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "dd38da62-bbcc-4728-af2b-1139632b0090",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "jax\n"
     ]
    }
   ],
   "source": [
    "print(ks.backend.backend())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "88f50597-0b72-47a6-9f05-0be224e91790",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from kgcnn.literature.GIN import make_model_edge as make_model\n",
    "from kgcnn.utils.plots  import plot_train_test_loss\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "c4ed1e84-e4e7-42f0-a42e-34fd413806a1",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:kgcnn.data.download:Checking and possibly downloading dataset with name FreeSolv\n",
      "INFO:kgcnn.data.download:Dataset directory located at C:\\Users\\patri\\.kgcnn\\datasets\n",
      "INFO:kgcnn.data.download:Dataset directory found. Done.\n",
      "INFO:kgcnn.data.download:Dataset found. Done.\n",
      "INFO:kgcnn.data.FreeSolv:Found SDF C:\\Users\\patri\\.kgcnn\\datasets\\FreeSolv\\SAMPL.sdf of pre-computed structures.\n",
      "INFO:kgcnn.data.FreeSolv:Read molecules from mol-file.\n",
      "INFO:kgcnn.data.FreeSolv: ... process molecules 0 from 642\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder Symbol found ['C', 'N', 'O', 'S', 'Cl', 'Br', 'P', 'F', 'I']\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder Hybridization found [rdkit.Chem.rdchem.HybridizationType.SP3, rdkit.Chem.rdchem.HybridizationType.SP2, rdkit.Chem.rdchem.HybridizationType.SP]\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder TotalDegree found [4, 3, 1, 2]\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder TotalNumHs found [3, 0, 1, 2, 4]\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder CIPCode found [None, 'S', 'R']\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder ChiralityPossible found [None, '1']\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder BondType found [rdkit.Chem.rdchem.BondType.SINGLE, rdkit.Chem.rdchem.BondType.DOUBLE, rdkit.Chem.rdchem.BondType.AROMATIC, rdkit.Chem.rdchem.BondType.TRIPLE]\n",
      "INFO:kgcnn.molecule.encoder:OneHotEncoder Stereo found [rdkit.Chem.rdchem.BondStereo.STEREONONE, rdkit.Chem.rdchem.BondStereo.STEREOZ, rdkit.Chem.rdchem.BondStereo.STEREOE]\n"
     ]
    }
   ],
   "source": [
    "from kgcnn.data.datasets.FreeSolvDataset import FreeSolvDataset\n",
    "dataset = FreeSolvDataset()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "000b1498-ddec-48bf-9e43-cc0b09b8eead",
   "metadata": {},
   "source": [
    "## Padded input"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "09e0cccd-3fcc-4a19-a612-d39d1de12ee5",
   "metadata": {},
   "source": [
    "With keeping the padded input in the disjoint representation and adding a dummy graph for the padded values which can be achieved with the `padded_disjoint` argument for the cast to disjoint layers in `\"cast_disjoint_kwargs\": {\"padded_disjoint\": True}` , the model can be jited.\n",
    "However, this will yield a significant number of padded values for node or edge attributes.\n",
    "For better performance it is better to padd the entire batch/disjoint node or edge tensors which requires a dataloader.\n",
    "\n",
    "Note that for the moment special layers like `BatchNormalization` must also passed the argument `padded_disjoint` but which is not yet fully supported."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6bbdd779-8297-4be9-a0d8-2674e5637c43",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_config = {\n",
    "    \"name\": \"GINE\",\n",
    "    \"inputs\": [\n",
    "        {\"shape\": (None, 41), \"name\": \"node_attributes\", \"dtype\": \"float32\"},\n",
    "        {\"shape\": (None, 11), \"name\": \"edge_attributes\", \"dtype\": \"float32\"},\n",
    "        {\"shape\": (None, 2), \"name\": \"edge_indices\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"total_nodes\", \"dtype\": \"int64\"},  # Or mask\n",
    "        {\"shape\": (), \"name\": \"total_edges\", \"dtype\": \"int64\"},  # Or mask\n",
    "    ],\n",
    "    \"input_tensor_type\": \"padded\",\n",
    "    \"cast_disjoint_kwargs\": {\"padded_disjoint\": True},\n",
    "    \"input_embedding\": None,  # deprecated\n",
    "    \"input_node_embedding\": {\"input_dim\": 95, \"output_dim\": 64},\n",
    "    \"input_edge_embedding\": {\"input_dim\": 10, \"output_dim\": 64},\n",
    "    \"gin_mlp\": {\"units\": [64, 64], \"use_bias\": True, \"activation\": [\"relu\", \"linear\"],\n",
    "                # At the moment we do not have a batch normalization for padded_disjoint\n",
    "                \"use_normalization\": False, \n",
    "                \"normalization_technique\": \"graph_batch\"},\n",
    "    \"gin_args\": {\"epsilon_learnable\": False},\n",
    "    \"depth\": 3, \"dropout\": 0.0, \"verbose\": 10,\n",
    "    \"last_mlp\": {\"use_bias\": [True, True, True], \"units\": [64, 64, 64],\n",
    "                 \"activation\": [\"relu\", \"relu\", \"linear\"]},\n",
    "    \"output_embedding\": 'graph',\n",
    "    \"output_to_tensor\": None,\n",
    "    \"output_scaling\": None,\n",
    "    \"output_tensor_type\": \"padded\",\n",
    "    \"output_mlp\": {\"use_bias\": True, \"units\": 1,\n",
    "                   \"activation\": \"linear\"}\n",
    "}\n",
    "inputs = model_config[\"inputs\"]\n",
    "outputs = {\"shape\": (1, ), \"name\": \"graph_labels\", \"dtype\": \"float32\"}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "30451b1f-d516-4ce5-bee6-6ae9d7441b3e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:kgcnn.data.FreeSolv:Property 'edge_attributes' is an empty list for graph '61'.\n",
      "INFO:kgcnn.data.FreeSolv:Property 'edge_attributes' is an empty list for graph '195'.\n",
      "INFO:kgcnn.data.FreeSolv:Property 'edge_attributes' is an empty list for graph '286'.\n",
      "INFO:kgcnn.data.FreeSolv:Property 'edge_indices' is an empty list for graph '61'.\n",
      "INFO:kgcnn.data.FreeSolv:Property 'edge_indices' is an empty list for graph '195'.\n",
      "INFO:kgcnn.data.FreeSolv:Property 'edge_indices' is an empty list for graph '286'.\n",
      "WARNING:kgcnn.data.FreeSolv:Found invalid graphs for properties. Removing graphs '[286 195  61]'.\n"
     ]
    }
   ],
   "source": [
    "dataset.map_list(method=\"count_nodes_and_edges\")\n",
    "dataset.clean(inputs)\n",
    "for i in range(len(dataset)):\n",
    "    if len(dataset[i][\"graph_labels\"].shape) < 1:\n",
    "        dataset[i][\"graph_labels\"] = np.expand_dims(dataset[i][\"graph_labels\"], axis=-1)\n",
    "train_indices, test_indices = train_test_split(np.arange(len(dataset)), test_size=0.2)\n",
    "\n",
    "x_train = dataset[train_indices].tensor(inputs)\n",
    "y_train = dataset[train_indices].tensor(outputs)\n",
    "x_test = dataset[test_indices].tensor(inputs)\n",
    "y_test = dataset[test_indices].tensor(outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "8cf506a5-55bb-4985-b3c8-516d73332406",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:kgcnn.models.utils:Updated model kwargs: '{'name': 'GINE', 'inputs': [{'shape': (None, 41), 'name': 'node_attributes', 'dtype': 'float32'}, {'shape': (None, 11), 'name': 'edge_attributes', 'dtype': 'float32'}, {'shape': (None, 2), 'name': 'edge_indices', 'dtype': 'int64'}, {'shape': (), 'name': 'total_nodes', 'dtype': 'int64'}, {'shape': (), 'name': 'total_edges', 'dtype': 'int64'}], 'input_tensor_type': 'padded', 'cast_disjoint_kwargs': {'padded_disjoint': True}, 'input_embedding': None, 'input_node_embedding': {'input_dim': 95, 'output_dim': 64}, 'input_edge_embedding': {'input_dim': 10, 'output_dim': 64}, 'gin_mlp': {'units': [64, 64], 'use_bias': True, 'activation': ['relu', 'linear'], 'use_normalization': False, 'normalization_technique': 'graph_batch'}, 'gin_args': {'epsilon_learnable': False}, 'depth': 3, 'dropout': 0.0, 'verbose': 10, 'last_mlp': {'use_bias': [True, True, True], 'units': [64, 64, 64], 'activation': ['relu', 'relu', 'linear']}, 'output_embedding': 'graph', 'output_to_tensor': None, 'output_scaling': None, 'output_tensor_type': 'padded', 'output_mlp': {'use_bias': True, 'units': 1, 'activation': 'linear'}}'.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Compiled with jit: True\n"
     ]
    }
   ],
   "source": [
    "model = make_model(**model_config)\n",
    "model.compile(\n",
    "    loss=\"mean_absolute_error\", \n",
    "    optimizer=ks.optimizers.Adam(learning_rate=1e-03),\n",
    "    jit_compile=True\n",
    ")\n",
    "print(\" Compiled with jit: %s\" % model._jit_compile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "a28b8310-910f-47e3-9e9e-898400ba07be",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\patri\\anaconda3\\envs\\gcnn_keras_test\\lib\\site-packages\\keras\\src\\backend\\jax\\core.py:76: UserWarning: Explicitly requested dtype int64 requested in asarray is not available, and will be truncated to dtype int32. To enable more dtypes, set the jax_enable_x64 configuration option or the JAX_ENABLE_X64 shell environment variable. See https://github.com/google/jax#current-gotchas for more.\n",
      "  return jnp.asarray(x, dtype=dtype)\n",
      "C:\\Users\\patri\\anaconda3\\envs\\gcnn_keras_test\\lib\\site-packages\\keras\\src\\backend\\jax\\numpy.py:224: UserWarning: Explicitly requested dtype int64 requested in arange is not available, and will be truncated to dtype int32. To enable more dtypes, set the jax_enable_x64 configuration option or the JAX_ENABLE_X64 shell environment variable. See https://github.com/google/jax#current-gotchas for more.\n",
      "  return jnp.arange(start, stop, step=step, dtype=dtype)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.34555432200431824\n"
     ]
    }
   ],
   "source": [
    "hist = model.fit(\n",
    "    x_train, y_train, \n",
    "    validation_data=(x_test, y_test),\n",
    "    epochs=100, \n",
    "    verbose=0\n",
    ")\n",
    "plot_train_test_loss([hist], dataset_name=\"FreeSolv\", model_name=\"GIN.make_model_edge\");\n",
    "print(hist.history[\"loss\"][-1])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b47652e0-30b0-49e4-9eaf-f5291ef65595",
   "metadata": {},
   "source": [
    "## Dataloader"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25961b4d-51eb-4f55-be06-f94175ccbbbe",
   "metadata": {},
   "source": [
    "The best way for training is to use a dataloader to generate a padded disjoint representation.\n",
    "One could recompile the model for each batch and run a fixed batch size. However, this is not tested and implemented.\n",
    "For now you can give the number of epochs and it will generate the maximum batch disjoint representation for the entire training process."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0d7dae3d-d1c3-432e-8948-3afa2b608fdf",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_config = {\n",
    "    \"name\": \"GINE\",\n",
    "    \"inputs\": [\n",
    "        {\"shape\": (41, ), \"name\": \"node_attributes\", \"dtype\": \"float32\"},\n",
    "        {\"shape\": (11, ), \"name\": \"edge_attributes\", \"dtype\": \"float32\"},\n",
    "        {\"shape\": (None, ), \"name\": \"edge_indices\", \"dtype\": \"int64\"},  # shape is (2, None)\n",
    "        {\"shape\": (), \"name\": \"batch_id_node\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"batch_id_edge\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"node_id\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"edge_id\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"count_nodes\", \"dtype\": \"int64\"},\n",
    "        {\"shape\": (), \"name\": \"count_edges\", \"dtype\": \"int64\"},\n",
    "    ],\n",
    "    \"input_tensor_type\": \"disjoint\",\n",
    "    \"cast_disjoint_kwargs\": {\"padded_disjoint\": True},\n",
    "    \"input_embedding\": None,  # deprecated\n",
    "    \"input_node_embedding\": {\"input_dim\": 95, \"output_dim\": 64},\n",
    "    \"input_edge_embedding\": {\"input_dim\": 10, \"output_dim\": 64},\n",
    "    \"gin_mlp\": {\"units\": [64, 64], \"use_bias\": True, \"activation\": [\"relu\", \"linear\"],\n",
    "                # At the moment we do not have a batch normalization for padded_disjoint\n",
    "                \"use_normalization\": False, \n",
    "                \"normalization_technique\": \"graph_batch\"},\n",
    "    \"gin_args\": {\"epsilon_learnable\": False},\n",
    "    \"depth\": 3, \"dropout\": 0.0, \"verbose\": 10,\n",
    "    \"last_mlp\": {\"use_bias\": [True, True, True], \"units\": [64, 64, 64],\n",
    "                 \"activation\": [\"relu\", \"relu\", \"linear\"]},\n",
    "    \"output_embedding\": 'graph',\n",
    "    \"output_to_tensor\": None,\n",
    "    \"output_scaling\": None,\n",
    "    \"output_tensor_type\": \"padded\",\n",
    "    \"output_mlp\": {\"use_bias\": True, \"units\": 1,\n",
    "                   \"activation\": \"linear\"}\n",
    "}\n",
    "inputs = model_config[\"inputs\"]\n",
    "outputs = {\"shape\": (1, ), \"name\": \"graph_labels\", \"dtype\": \"float32\"}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f98401c3-281d-4e7b-a079-b2ff3795b5ad",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:kgcnn.models.utils:Updated model kwargs: '{'name': 'GINE', 'inputs': [{'shape': (41,), 'name': 'node_attributes', 'dtype': 'float32'}, {'shape': (11,), 'name': 'edge_attributes', 'dtype': 'float32'}, {'shape': (None,), 'name': 'edge_indices', 'dtype': 'int64'}, {'shape': (), 'name': 'batch_id_node', 'dtype': 'int64'}, {'shape': (), 'name': 'batch_id_edge', 'dtype': 'int64'}, {'shape': (), 'name': 'node_id', 'dtype': 'int64'}, {'shape': (), 'name': 'edge_id', 'dtype': 'int64'}, {'shape': (), 'name': 'count_nodes', 'dtype': 'int64'}, {'shape': (), 'name': 'count_edges', 'dtype': 'int64'}], 'input_tensor_type': 'disjoint', 'cast_disjoint_kwargs': {'padded_disjoint': True}, 'input_embedding': None, 'input_node_embedding': {'input_dim': 95, 'output_dim': 64}, 'input_edge_embedding': {'input_dim': 10, 'output_dim': 64}, 'gin_mlp': {'units': [64, 64], 'use_bias': True, 'activation': ['relu', 'linear'], 'use_normalization': False, 'normalization_technique': 'graph_batch'}, 'gin_args': {'epsilon_learnable': False}, 'depth': 3, 'dropout': 0.0, 'verbose': 10, 'last_mlp': {'use_bias': [True, True, True], 'units': [64, 64, 64], 'activation': ['relu', 'relu', 'linear']}, 'output_embedding': 'graph', 'output_to_tensor': None, 'output_scaling': None, 'output_tensor_type': 'padded', 'output_mlp': {'use_bias': True, 'units': 1, 'activation': 'linear'}}'.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Compiled with jit: True\n"
     ]
    }
   ],
   "source": [
    "model = make_model(**model_config)\n",
    "model.compile(\n",
    "    loss=\"mean_absolute_error\", \n",
    "    optimizer=ks.optimizers.Adam(learning_rate=1e-03),\n",
    "    jit_compile=True\n",
    ")\n",
    "print(\" Compiled with jit: %s\" % model._jit_compile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "2c233f60-65e7-4896-8269-840f51cf3110",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:kgcnn.io.loader:Max of graph: {0: 24, 1: 50, 2: 50, 3: None, 4: None, 5: None, 6: None, 7: None, 8: None}.\n",
      "INFO:kgcnn.io.loader:Padded max of disjoint: [11.53125, 23.375, 23.375, None, None, None, None, None, None].\n",
      "INFO:kgcnn.io.loader:Max of graph: {0: 22, 1: 46, 2: 46, 3: None, 4: None, 5: None, 6: None, 7: None, 8: None}.\n",
      "INFO:kgcnn.io.loader:Padded max of disjoint: [10.25, 20.0625, 20.0625, None, None, None, None, None, None].\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.34142956137657166\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "# For train\n",
    "dataset_train = dataset[train_indices]\n",
    "dataloader_train_x = dataset_train.tf_dataset_disjoint(\n",
    "    inputs=inputs, \n",
    "    batch_size=32, \n",
    "    assignment_to_id=[0, 1, 1],\n",
    "    assignment_of_indices=[None, None, 0],\n",
    "    pos_batch_id=[3,4],\n",
    "    pos_subgraph_id=[5,6],\n",
    "    pos_count=[7,8],\n",
    "    epochs=100,\n",
    "    padded_disjoint=True\n",
    ")\n",
    "dataloader_train_y = dataset_train.tf_dataset_disjoint(\n",
    "    inputs=outputs, \n",
    "    batch_size=32, \n",
    "    epochs=100,\n",
    "    padded_disjoint=False\n",
    ")\n",
    "dataloader_train = tf.data.Dataset.zip((dataloader_train_x, dataloader_train_y))\n",
    "# For test\n",
    "dataset_test = dataset[test_indices]\n",
    "dataloader_test_x = dataset_test.tf_dataset_disjoint(\n",
    "    inputs=inputs, \n",
    "    batch_size=32, \n",
    "    assignment_to_id=[0, 1, 1],\n",
    "    assignment_of_indices=[None, None, 0],\n",
    "    pos_batch_id=[3,4],\n",
    "    pos_subgraph_id=[5,6],\n",
    "    pos_count=[7,8],\n",
    "    epochs=100,\n",
    "    padded_disjoint=True\n",
    ")\n",
    "dataloader_test_y = dataset_test.tf_dataset_disjoint(\n",
    "    inputs=outputs, \n",
    "    batch_size=32, \n",
    "    epochs=100,\n",
    "    padded_disjoint=False\n",
    ")\n",
    "dataloader_test = tf.data.Dataset.zip((dataloader_test_x, dataloader_test_y))\n",
    "hist = model.fit(\n",
    "    dataloader_train, \n",
    "    validation_data=dataloader_test, \n",
    "    epochs=100,\n",
    "    verbose=0\n",
    ")\n",
    "plot_train_test_loss([hist], dataset_name=\"FreeSolv\", model_name=\"GIN.make_model_edge\");\n",
    "print(hist.history[\"loss\"][-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "02333c75-0703-4543-8da8-c6bb0f0184be",
   "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": "3.10.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}