{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Implementing your own lifting" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "In this tutorial we show how you can implement your own lifting and test it on a dataset. \n", "\n", "This particular example uses the MUTAG dataset. The lifting for this example is similar to the SimplicialCliqueLifting but discards the cliques that are bigger than the maximum simplices we want to consider.\n", "\n", "We test this lifting using the SCN2 model from `TopoModelX`." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Table of contents\n", " [1. Imports](##sec1)\n", "\n", " [2. Configurations and utilities](##sec2)\n", "\n", " [3. Defining the lifting](##sec2)\n", "\n", " [4. Loading the data](##sec3)\n", "\n", " [5. Model initialization](##sec4)\n", "\n", " [6. Training](##sec5)\n", "\n", " [7. Testing the model](##sec6)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 1. Imports " ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "from itertools import combinations\n", "from typing import Any\n", "\n", "import lightning as pl\n", "import networkx as nx\n", "import torch\n", "import torch_geometric\n", "from omegaconf import OmegaConf\n", "from topomodelx.nn.simplicial.scn2 import SCN2\n", "from toponetx.classes import SimplicialComplex\n", "\n", "from topobenchmarkx.data.loaders import GraphLoader\n", "from topobenchmarkx.data.preprocessor import PreProcessor\n", "from topobenchmarkx.dataloader import TBXDataloader\n", "from topobenchmarkx.evaluator import TBXEvaluator\n", "from topobenchmarkx.loss import TBXLoss\n", "from topobenchmarkx.model import TBXModel\n", "from topobenchmarkx.nn.encoders import AllCellFeatureEncoder\n", "from topobenchmarkx.nn.readouts import PropagateSignalDown\n", "from topobenchmarkx.nn.wrappers.simplicial import SCNWrapper\n", "from topobenchmarkx.optimizer import TBXOptimizer\n", "from topobenchmarkx.transforms.liftings.graph2simplicial import (\n", " Graph2SimplicialLifting,\n", ")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 2. Configurations and utilities " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Configurations can be specified using yaml files or directly specified in your code like in this example. To keep the notebook clean here we already define the configuration for the lifting, which is defined later in the notebook." ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "loader_config = {\n", " \"data_domain\": \"graph\",\n", " \"data_type\": \"TUDataset\",\n", " \"data_name\": \"MUTAG\",\n", " \"data_dir\": \"./data/MUTAG/\",\n", "}\n", "\n", "transform_config = { \"clique_lifting\":\n", " {\"_target_\": \"__main__.SimplicialCliquesLEQLifting\",\n", " \"transform_name\": \"SimplicialCliquesLEQLifting\",\n", " \"transform_type\": \"lifting\",\n", " \"complex_dim\": 3,}\n", "}\n", "\n", "split_config = {\n", " \"learning_setting\": \"inductive\",\n", " \"split_type\": \"k-fold\",\n", " \"data_seed\": 0,\n", " \"data_split_dir\": \"./data/MUTAG/splits/\",\n", " \"k\": 10,\n", "}\n", "\n", "in_channels = 7\n", "out_channels = 2\n", "dim_hidden = 128\n", "\n", "wrapper_config = {\n", " \"out_channels\": dim_hidden,\n", " \"num_cell_dimensions\": 3,\n", "}\n", "\n", "readout_config = {\n", " \"readout_name\": \"PropagateSignalDown\",\n", " \"num_cell_dimensions\": 1,\n", " \"hidden_dim\": dim_hidden,\n", " \"out_channels\": out_channels,\n", " \"task_level\": \"graph\",\n", " \"pooling_type\": \"sum\",\n", "}\n", "\n", "loss_config = {\"task\": \"classification\", \"loss_type\": \"cross_entropy\"}\n", "\n", "evaluator_config = {\"task\": \"classification\",\n", " \"num_classes\": out_channels,\n", " \"metrics\": [\"accuracy\", \"precision\", \"recall\"]}\n", "\n", "optimizer_config = {\"optimizer_id\": \"Adam\",\n", " \"parameters\":\n", " {\"lr\": 0.001,\"weight_decay\": 0.0005}\n", " }\n", "\n", "\n", "loader_config = OmegaConf.create(loader_config)\n", "transform_config = OmegaConf.create(transform_config)\n", "split_config = OmegaConf.create(split_config)\n", "wrapper_config = OmegaConf.create(wrapper_config)\n", "readout_config = OmegaConf.create(readout_config)\n", "loss_config = OmegaConf.create(loss_config)\n", "evaluator_config = OmegaConf.create(evaluator_config)\n", "optimizer_config = OmegaConf.create(optimizer_config)" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "def wrapper(**factory_kwargs):\n", " def factory(backbone):\n", " return SCNWrapper(backbone, **factory_kwargs)\n", " return factory" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 3. Defining the lifting " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Here we define the lifting we intend on using. The `SimplicialCliquesLEQLifting` finds the cliques that have a number of nodes less or equal to the maximum simplices we want to consider and creates simplices from them. The configuration for the lifting was already defined with the other configurations." ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "class SimplicialCliquesLEQLifting(Graph2SimplicialLifting):\n", " r\"\"\"Lifts graphs to simplicial complex domain by identifying the cliques as k-simplices. Only the cliques with size smaller or equal to the max complex dimension are considered.\n", " \n", " Args:\n", " kwargs (optional): Additional arguments for the class.\n", " \"\"\"\n", " def __init__(self, **kwargs):\n", " super().__init__(**kwargs)\n", "\n", " def lift_topology(self, data: torch_geometric.data.Data) -> dict:\n", " r\"\"\"Lifts the topology of a graph to a simplicial complex by identifying the cliques as k-simplices. Only the cliques with size smaller or equal to the max complex dimension are considered.\n", "\n", " Args:\n", " data (torch_geometric.data.Data): The input data to be lifted.\n", " Returns:\n", " dict: The lifted topology.\n", " \"\"\"\n", " graph = self._generate_graph_from_data(data)\n", " simplicial_complex = SimplicialComplex(graph)\n", " cliques = nx.find_cliques(graph)\n", " \n", " simplices: list[set[tuple[Any, ...]]] = [set() for _ in range(2, self.complex_dim + 1)]\n", " for clique in cliques:\n", " if len(clique) <= self.complex_dim + 1:\n", " for i in range(2, self.complex_dim + 1):\n", " for c in combinations(clique, i + 1):\n", " simplices[i - 2].add(tuple(c))\n", "\n", " for set_k_simplices in simplices:\n", " simplicial_complex.add_simplices_from(list(set_k_simplices))\n", "\n", " return self._get_lifted_topology(simplicial_complex, graph)\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 4. Loading the data " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "In this example we use the MUTAG dataset." ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "from topobenchmarkx.transforms import TRANSFORMS\n", "\n", "TRANSFORMS[\"SimplicialCliquesLEQLifting\"] = SimplicialCliquesLEQLifting" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Transform parameters are the same, using existing data_dir: ./data/MUTAG/MUTAG/clique_lifting/458544608\n" ] } ], "source": [ "graph_loader = GraphLoader(loader_config)\n", "\n", "dataset, dataset_dir = graph_loader.load()\n", "\n", "preprocessor = PreProcessor(dataset, dataset_dir, transform_config)\n", "dataset_train, dataset_val, dataset_test = preprocessor.load_dataset_splits(split_config)\n", "datamodule = TBXDataloader(dataset_train, dataset_val, dataset_test, batch_size=32)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 5. Model initialization " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "We can create the backbone by instantiating the SCN2 model form TopoModelX. Then the `SCNWrapper` and the `TBXModel` take care of the rest." ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [], "source": [ "backbone = SCN2(in_channels_0=dim_hidden,in_channels_1=dim_hidden,in_channels_2=dim_hidden)\n", "backbone_wrapper = wrapper(**wrapper_config)\n", "\n", "readout = PropagateSignalDown(**readout_config)\n", "loss = TBXLoss(**loss_config)\n", "feature_encoder = AllCellFeatureEncoder(in_channels=[in_channels, in_channels, in_channels], out_channels=dim_hidden)\n", "\n", "evaluator = TBXEvaluator(**evaluator_config)\n", "optimizer = TBXOptimizer(**optimizer_config)" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": [ "model = TBXModel(backbone=backbone,\n", " backbone_wrapper=backbone_wrapper,\n", " readout=readout,\n", " loss=loss,\n", " feature_encoder=feature_encoder,\n", " evaluator=evaluator,\n", " optimizer=optimizer,\n", " compile=False,)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 6. Training " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Now we can use the `lightning` trainer to train the model. We are prompted to connet a Wandb account to monitor training, but we can also obtain the final training metrics from the trainer directly." ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "GPU available: True (mps), used: False\n", "TPU available: False, using: 0 TPU cores\n", "IPU available: False, using: 0 IPUs\n", "HPU available: False, using: 0 HPUs\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/trainer/setup.py:187: GPU available but not used. You can set it by doing `Trainer(accelerator='gpu')`.\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/trainer/connectors/logger_connector/logger_connector.py:75: Starting from v1.9.0, `tensorboardX` has been removed as a dependency of the `lightning.pytorch` package, due to potential conflicts with other packages in the ML ecosystem. For this reason, `logger=True` will use `CSVLogger` as the default logger, unless the `tensorboard` or `tensorboardX` packages are found. Please `pip install lightning[extra]` or one of them to enable TensorBoard support by default\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/utilities/parsing.py:44: attribute 'backbone_wrapper' removed from hparams because it cannot be pickled\n", "\n", " | Name | Type | Params\n", "----------------------------------------------------------\n", "0 | feature_encoder | AllCellFeatureEncoder | 53.8 K\n", "1 | backbone | SCNWrapper | 99.1 K\n", "2 | readout | PropagateSignalDown | 258 \n", "3 | val_acc_best | MeanMetric | 0 \n", "----------------------------------------------------------\n", "153 K Trainable params\n", "0 Non-trainable params\n", "153 K Total params\n", "0.612 Total estimated model params size (MB)\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=13` in the `DataLoader` to improve performance.\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/torchmetrics/utilities/prints.py:43: UserWarning: The ``compute`` method of metric MulticlassAccuracy was called before the ``update`` method which may lead to errors, as metric states have not yet been updated.\n", " warnings.warn(*args, **kwargs) # noqa: B028\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/torchmetrics/utilities/prints.py:43: UserWarning: The ``compute`` method of metric MulticlassPrecision was called before the ``update`` method which may lead to errors, as metric states have not yet been updated.\n", " warnings.warn(*args, **kwargs) # noqa: B028\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/torchmetrics/utilities/prints.py:43: UserWarning: The ``compute`` method of metric MulticlassRecall was called before the ``update`` method which may lead to errors, as metric states have not yet been updated.\n", " warnings.warn(*args, **kwargs) # noqa: B028\n", "/Users/gbg141/Documents/TopoProjectX/TopoBenchmarkX/topobenchmarkx/nn/wrappers/simplicial/scn_wrapper.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /Users/runner/work/pytorch/pytorch/pytorch/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)\n", " normalized_matrix = diag_matrix @ (matrix @ diag_matrix)\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=13` in the `DataLoader` to improve performance.\n", "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/loops/fit_loop.py:298: The number of training batches (6) is smaller than the logging interval Trainer(log_every_n_steps=50). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.\n", "`Trainer.fit` stopped: `max_epochs=50` reached.\n" ] } ], "source": [ "# Increase the number of epochs to get better results\n", "trainer = pl.Trainer(max_epochs=50, accelerator=\"cpu\", enable_progress_bar=False)\n", "\n", "trainer.fit(model, datamodule)\n", "train_metrics = trainer.callback_metrics" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " Training metrics\n", " --------------------------\n", "train/accuracy: 0.7870\n", "train/precision: 0.7628\n", "train/recall: 0.7531\n", "val/loss: 0.6316\n", "val/accuracy: 0.7895\n", "val/precision: 0.7750\n", "val/recall: 0.7115\n", "train/loss: 0.4959\n" ] } ], "source": [ "print(' Training metrics\\n', '-'*26)\n", "for key in train_metrics:\n", " print('{:<21s} {:>5.4f}'.format(key+':', train_metrics[key].item()))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 7. Testing the model " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Finally, we can test the model and obtain the results." ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "/opt/miniconda3/envs/topox/lib/python3.11/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=13` in the `DataLoader` to improve performance.\n" ] }, { "data": { "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
       "┃        Test metric        ┃       DataLoader 0        ┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩\n",
       "│       test/accuracy       │    0.7894737124443054     │\n",
       "│         test/loss         │     0.631595253944397     │\n",
       "│      test/precision       │    0.7749999761581421     │\n",
       "│        test/recall        │    0.7115384340286255     │\n",
       "└───────────────────────────┴───────────────────────────┘\n",
       "
\n" ], "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1m Test metric \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m DataLoader 0 \u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩\n", "│\u001b[36m \u001b[0m\u001b[36m test/accuracy \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m 0.7894737124443054 \u001b[0m\u001b[35m \u001b[0m│\n", "│\u001b[36m \u001b[0m\u001b[36m test/loss \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m 0.631595253944397 \u001b[0m\u001b[35m \u001b[0m│\n", "│\u001b[36m \u001b[0m\u001b[36m test/precision \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m 0.7749999761581421 \u001b[0m\u001b[35m \u001b[0m│\n", "│\u001b[36m \u001b[0m\u001b[36m test/recall \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m 0.7115384340286255 \u001b[0m\u001b[35m \u001b[0m│\n", "└───────────────────────────┴───────────────────────────┘\n" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "trainer.test(model, datamodule)\n", "test_metrics = trainer.callback_metrics" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " Testing metrics\n", " -------------------------\n", "test/loss: 0.6316\n", "test/accuracy: 0.7895\n", "test/precision: 0.7750\n", "test/recall: 0.7115\n" ] } ], "source": [ "print(' Testing metrics\\n', '-'*25)\n", "for key in test_metrics:\n", " print('{:<20s} {:>5.4f}'.format(key+':', test_metrics[key].item()))" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "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": "3.11.3" }, "orig_nbformat": 4 }, "nbformat": 4, "nbformat_minor": 2 }