Implementing your own lifting#
In this tutorial we show how you can implement your own lifting and test it on a dataset.
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.
We test this lifting using the SCN2 model from TopoModelX.
Table of contents#
1. Imports
2. Configurations and utilities
3. Defining the lifting
4. Loading the data
5. Model initialization
6. Training
7. Testing the model
1. Imports#
In [1]:
from itertools import combinations
from typing import Any
import lightning as pl
import networkx as nx
import torch
import torch_geometric
from omegaconf import OmegaConf
from topomodelx.nn.simplicial.scn2 import SCN2
from toponetx.classes import SimplicialComplex
from topobenchmarkx.data.loaders import GraphLoader
from topobenchmarkx.data.preprocessor import PreProcessor
from topobenchmarkx.dataloader import TBXDataloader
from topobenchmarkx.evaluator import TBXEvaluator
from topobenchmarkx.loss import TBXLoss
from topobenchmarkx.model import TBXModel
from topobenchmarkx.nn.encoders import AllCellFeatureEncoder
from topobenchmarkx.nn.readouts import PropagateSignalDown
from topobenchmarkx.nn.wrappers.simplicial import SCNWrapper
from topobenchmarkx.optimizer import TBXOptimizer
from topobenchmarkx.transforms.liftings.graph2simplicial import (
Graph2SimplicialLifting,
)
2. Configurations and utilities#
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.
In [2]:
loader_config = {
"data_domain": "graph",
"data_type": "TUDataset",
"data_name": "MUTAG",
"data_dir": "./data/MUTAG/",
}
transform_config = { "clique_lifting":
{"_target_": "__main__.SimplicialCliquesLEQLifting",
"transform_name": "SimplicialCliquesLEQLifting",
"transform_type": "lifting",
"complex_dim": 3,}
}
split_config = {
"learning_setting": "inductive",
"split_type": "k-fold",
"data_seed": 0,
"data_split_dir": "./data/MUTAG/splits/",
"k": 10,
}
in_channels = 7
out_channels = 2
dim_hidden = 128
wrapper_config = {
"out_channels": dim_hidden,
"num_cell_dimensions": 3,
}
readout_config = {
"readout_name": "PropagateSignalDown",
"num_cell_dimensions": 1,
"hidden_dim": dim_hidden,
"out_channels": out_channels,
"task_level": "graph",
"pooling_type": "sum",
}
loss_config = {"task": "classification", "loss_type": "cross_entropy"}
evaluator_config = {"task": "classification",
"num_classes": out_channels,
"metrics": ["accuracy", "precision", "recall"]}
optimizer_config = {"optimizer_id": "Adam",
"parameters":
{"lr": 0.001,"weight_decay": 0.0005}
}
loader_config = OmegaConf.create(loader_config)
transform_config = OmegaConf.create(transform_config)
split_config = OmegaConf.create(split_config)
wrapper_config = OmegaConf.create(wrapper_config)
readout_config = OmegaConf.create(readout_config)
loss_config = OmegaConf.create(loss_config)
evaluator_config = OmegaConf.create(evaluator_config)
optimizer_config = OmegaConf.create(optimizer_config)
In [3]:
def wrapper(**factory_kwargs):
def factory(backbone):
return SCNWrapper(backbone, **factory_kwargs)
return factory
3. Defining the lifting#
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.
In [4]:
class SimplicialCliquesLEQLifting(Graph2SimplicialLifting):
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.
Args:
kwargs (optional): Additional arguments for the class.
"""
def __init__(self, **kwargs):
super().__init__(**kwargs)
def lift_topology(self, data: torch_geometric.data.Data) -> dict:
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.
Args:
data (torch_geometric.data.Data): The input data to be lifted.
Returns:
dict: The lifted topology.
"""
graph = self._generate_graph_from_data(data)
simplicial_complex = SimplicialComplex(graph)
cliques = nx.find_cliques(graph)
simplices: list[set[tuple[Any, ...]]] = [set() for _ in range(2, self.complex_dim + 1)]
for clique in cliques:
if len(clique) <= self.complex_dim + 1:
for i in range(2, self.complex_dim + 1):
for c in combinations(clique, i + 1):
simplices[i - 2].add(tuple(c))
for set_k_simplices in simplices:
simplicial_complex.add_simplices_from(list(set_k_simplices))
return self._get_lifted_topology(simplicial_complex, graph)
4. Loading the data#
In this example we use the MUTAG dataset.
In [5]:
from topobenchmarkx.transforms import TRANSFORMS
TRANSFORMS["SimplicialCliquesLEQLifting"] = SimplicialCliquesLEQLifting
In [6]:
graph_loader = GraphLoader(loader_config)
dataset, dataset_dir = graph_loader.load()
preprocessor = PreProcessor(dataset, dataset_dir, transform_config)
dataset_train, dataset_val, dataset_test = preprocessor.load_dataset_splits(split_config)
datamodule = TBXDataloader(dataset_train, dataset_val, dataset_test, batch_size=32)
Transform parameters are the same, using existing data_dir: ./data/MUTAG/MUTAG/clique_lifting/458544608
5. Model initialization#
We can create the backbone by instantiating the SCN2 model form TopoModelX. Then the SCNWrapper and the TBXModel take care of the rest.
In [7]:
backbone = SCN2(in_channels_0=dim_hidden,in_channels_1=dim_hidden,in_channels_2=dim_hidden)
backbone_wrapper = wrapper(**wrapper_config)
readout = PropagateSignalDown(**readout_config)
loss = TBXLoss(**loss_config)
feature_encoder = AllCellFeatureEncoder(in_channels=[in_channels, in_channels, in_channels], out_channels=dim_hidden)
evaluator = TBXEvaluator(**evaluator_config)
optimizer = TBXOptimizer(**optimizer_config)
In [8]:
model = TBXModel(backbone=backbone,
backbone_wrapper=backbone_wrapper,
readout=readout,
loss=loss,
feature_encoder=feature_encoder,
evaluator=evaluator,
optimizer=optimizer,
compile=False,)
6. Training#
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.
In [9]:
# Increase the number of epochs to get better results
trainer = pl.Trainer(max_epochs=50, accelerator="cpu", enable_progress_bar=False)
trainer.fit(model, datamodule)
train_metrics = trainer.callback_metrics
GPU available: True (mps), used: False
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
/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')`.
/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
/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
| Name | Type | Params
----------------------------------------------------------
0 | feature_encoder | AllCellFeatureEncoder | 53.8 K
1 | backbone | SCNWrapper | 99.1 K
2 | readout | PropagateSignalDown | 258
3 | val_acc_best | MeanMetric | 0
----------------------------------------------------------
153 K Trainable params
0 Non-trainable params
153 K Total params
0.612 Total estimated model params size (MB)
/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.
/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.
warnings.warn(*args, **kwargs) # noqa: B028
/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.
warnings.warn(*args, **kwargs) # noqa: B028
/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.
warnings.warn(*args, **kwargs) # noqa: B028
/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.)
normalized_matrix = diag_matrix @ (matrix @ diag_matrix)
/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.
/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.
`Trainer.fit` stopped: `max_epochs=50` reached.
In [10]:
print(' Training metrics\n', '-'*26)
for key in train_metrics:
print('{:<21s} {:>5.4f}'.format(key+':', train_metrics[key].item()))
Training metrics
--------------------------
train/accuracy: 0.7870
train/precision: 0.7628
train/recall: 0.7531
val/loss: 0.6316
val/accuracy: 0.7895
val/precision: 0.7750
val/recall: 0.7115
train/loss: 0.4959
7. Testing the model#
Finally, we can test the model and obtain the results.
In [11]:
trainer.test(model, datamodule)
test_metrics = trainer.callback_metrics
/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.
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ ┃ Test metric ┃ DataLoader 0 ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ │ test/accuracy │ 0.7894737124443054 │ │ test/loss │ 0.631595253944397 │ │ test/precision │ 0.7749999761581421 │ │ test/recall │ 0.7115384340286255 │ └───────────────────────────┴───────────────────────────┘
In [12]:
print(' Testing metrics\n', '-'*25)
for key in test_metrics:
print('{:<20s} {:>5.4f}'.format(key+':', test_metrics[key].item()))
Testing metrics
-------------------------
test/loss: 0.6316
test/accuracy: 0.7895
test/precision: 0.7750
test/recall: 0.7115
In [ ]: