statistics dataset script

.dockerignore

@@ -0,0 +1,3 @@
+/logs
+/datasets/graph
+.ruff_cache

Dockerfile

@@ -0,0 +1,16 @@
+FROM python:3.11.3
+
+WORKDIR /TopoBenchmarkX
+
+COPY . .
+
+RUN pip install --upgrade pip
+
+RUN pip install -e '.[all]'
+RUN pip install --no-dependencies git+https://github.com/pyt-team/TopoNetX.git
+RUN pip install --no-dependencies git+https://github.com/pyt-team/TopoModelX.git
+RUN pip install torch==2.0.1 --extra-index-url https://download.pytorch.org/whl/cu115
+RUN pip install torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.0.1+cu115.html
+RUN pip install torch-cluster -f https://data.pyg.org/whl/torch-2.0.0+cu115.html
+#RUN pip install lightning>=2.0.0
+#RUN pip install numpy pre-commit jupyterlab notebook ipykernel

configs/model/hypergraph/unignn.yaml

@@ -1,7 +1,7 @@
 _target_: topobenchmarkx.models.TopologicalNetworkModule
 
 model_name: unignn2
-mode_domain: hypergraph
+model_domain: hypergraph
 
 feature_encoder:
   _target_: topobenchmarkx.models.encoders.${model.feature_encoder.encoder_name}

configs/model/hypergraph/unignn2.yaml

@@ -1,7 +1,7 @@
 _target_: topobenchmarkx.models.TopologicalNetworkModule
 
 model_name: unignn2
-mode_domain: hypergraph
+model_domain: hypergraph
 
 feature_encoder:
   _target_: topobenchmarkx.models.encoders.${model.feature_encoder.encoder_name}

configs/model/simplicial/scn.yaml

@@ -1,7 +1,7 @@
 _target_: topobenchmarkx.models.TopologicalNetworkModule
 
 mdoel_name: scn 
-model_type: simplicial
+model_domain: simplicial
 
 feature_encoder:
   _target_: topobenchmarkx.models.encoders.${model.feature_encoder.encoder_name}

configs/train.yaml

@@ -4,8 +4,8 @@
 # order of defaults determines the order in which configs override each other
 defaults:
   - _self_
-  - dataset: NCI1 #us_country_demos
-  - model: graph/gcn #hypergraph/unignn2 #allsettransformer
+  - dataset: cocitation_cora #us_country_demos
+  - model: simplicial/scn #hypergraph/unignn2 #allsettransformer
   - evaluator: default
   - callbacks: default
   - logger: wandb # set logger here or use command line (e.g. `python train.py logger=tensorboard`)

env.sh

@@ -0,0 +1,21 @@
+# #!/bin/bash
+
+#conda create -n topoxx python=3.11.3
+#conda activate topoxx
+
+pip install --upgrade pip
+pip install -e '.[all]'
+
+pip install git+https://github.com/pyt-team/TopoNetX.git
+pip install git+https://github.com/pyt-team/TopoModelX.git
+pip install git+https://github.com/pyt-team/TopoEmbedX.git
+
+CUDA="cu117" # if available, select the CUDA version suitable for your system
+             # e.g. cpu, cu102, cu111, cu113, cu115
+pip install torch==2.0.1 --extra-index-url https://download.pytorch.org/whl/${CUDA}
+pip install torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.0.1+${CUDA}.html
+pip install torch-cluster -f https://data.pyg.org/whl/torch-2.0.0+${CUDA}.html
+
+pytest
+
+pre-commit install

tables/dataset_statistics.csv

@@ -0,0 +1,7 @@
+,num_hyperedges,zero_cell,one_cell,two_cell,three_cell,dataset,domain
+0,0,3224,9483,6266,0,US-county-demos,cell
+1,0,2708,5278,2648,0,Cora,cell
+2,0,3327,4552,1663,0,citeseer,cell
+3,0,19717,44324,23605,0,PubMed,cell
+4,0,277864,298985,33121,0,ZINC,cell
+5,0,22662,32927,10266,0,roman_empire,cell

topobenchmarkx/dataset_statistics.py

@@ -0,0 +1,224 @@
+import random
+from typing import Any
+
+import hydra
+import lightning as L
+import numpy as np
+import rootutils
+
+rootutils.setup_root(__file__, indicator=".project-root", pythonpath=True)
+import torch
+from lightning import Callback, LightningModule, Trainer
+from lightning.pytorch.loggers import Logger
+from omegaconf import DictConfig, OmegaConf
+
+from topobenchmarkx.data.dataloaders import DefaultDataModule
+from topobenchmarkx.utils import (
+    RankedLogger,
+    extras,
+    get_metric_value,
+    instantiate_callbacks,
+    instantiate_loggers,
+    log_hyperparameters,
+    task_wrapper,
+)
+
+from topobenchmarkx.utils.config_resolvers import (
+    get_default_transform,
+    get_monitor_metric,
+    get_monitor_mode,
+    infer_in_channels,
+    infere_list_length,
+)
+import pandas as pd
+import os
+# ------------------------------------------------------------------------------------ #
+# the setup_root above is equivalent to:
+# - adding project root dir to PYTHONPATH
+#       (so you don't need to force user to install project as a package)
+#       (necessary before importing any local modules e.g. `from src import utils`)
+# - setting up PROJECT_ROOT environment variable
+#       (which is used as a base for paths in "configs/paths/default.yaml")
+#       (this way all filepaths are the same no matter where you run the code)
+# - loading environment variables from ".env" in root dir
+#
+# you can remove it if you:
+# 1. either install project as a package or move entry files to project root dir
+# 2. set `root_dir` to "." in "configs/paths/default.yaml"
+#
+# more info: https://github.com/ashleve/rootutils
+# ------------------------------------------------------------------------------------ #
+
+
+OmegaConf.register_new_resolver("get_default_transform", get_default_transform)
+OmegaConf.register_new_resolver("get_monitor_metric", get_monitor_metric)
+OmegaConf.register_new_resolver("get_monitor_mode", get_monitor_mode)
+OmegaConf.register_new_resolver("infer_in_channels", infer_in_channels)
+OmegaConf.register_new_resolver("infere_list_length", infere_list_length)
+OmegaConf.register_new_resolver(
+    "parameter_multiplication", lambda x, y: int(int(x) * int(y))
+)
+
+torch.set_num_threads(1)
+log = RankedLogger(__name__, rank_zero_only=True)
+
+
+
+def train(cfg: DictConfig) -> tuple[dict[str, Any], dict[str, Any]]:
+    """Trains the model. Can additionally evaluate on a testset, using best
+    weights obtained during training.
+
+    This method is wrapped in optional @task_wrapper decorator, that controls
+    the behavior during failure. Useful for multiruns, saving info about the
+    crash, etc.
+
+    :param cfg: A DictConfig configuration composed by Hydra.
+    :return: A tuple with metrics and dict with all instantiated objects.
+    """
+
+    # Set seed for random number generators in pytorch, numpy and python.random
+    # if cfg.get("seed"):
+    L.seed_everything(cfg.seed, workers=True)
+    # Seed for torch
+    torch.manual_seed(cfg.seed)
+    # Seed for numpy
+    np.random.seed(cfg.seed)
+    # Seed for python random
+    random.seed(cfg.seed)
+
+    if cfg.model.model_domain == "cell":
+        cfg.dataset.transforms.graph2cell_lifting.max_cell_length=1000
+
+    # Instantiate and load dataset
+    dataset = hydra.utils.instantiate(cfg.dataset, _recursive_=False)
+    dataset = dataset.load()
+
+    one_graph_flag = True
+    if cfg.dataset.parameters.batch_size != 1:
+        cfg.dataset.parameters.batch_size != 1
+        one_graph_flag = False
+
+
+    log.info(f"Instantiating datamodule <{cfg.dataset._target_}>")
+
+    if cfg.dataset.parameters.task_level == "node":
+        datamodule = DefaultDataModule(dataset_train=dataset)
+
+    elif cfg.dataset.parameters.task_level == "graph":
+        datamodule = DefaultDataModule(
+            dataset_train=dataset[0],
+            dataset_val=dataset[1],
+            dataset_test=dataset[2],
+            batch_size=cfg.dataset.parameters.batch_size,
+        )
+
+    else:
+        raise ValueError("Invalid task_level")
+
+    if one_graph_flag == True:
+        dataloaders = [datamodule.train_dataloader()]
+    else:
+        dataloaders = [datamodule.train_dataloader(), datamodule.val_dataloader(), datamodule.test_dataloader()]
+
+    dict_collector = {
+        "num_hyperedges": 0,
+        "zero_cell": 0,
+        "one_cell": 0,
+        "two_cell": 0,
+        "three_cell": 0,
+    }
+
+    for loader in dataloaders:
+        for batch in loader:
+            if cfg.model.model_domain == "hypergraph":
+                dict_collector["zero_cell"] += batch.x.shape[0]
+                dict_collector["num_hyperedges"] += batch.x_hyperedges.shape[0]
+
+            elif cfg.model.model_domain == "simplicial":
+                dict_collector["zero_cell"] += batch.x_0.shape[0]
+                dict_collector["one_cell"] +=batch.x_1.shape[0]
+                dict_collector["two_cell"] +=batch.x_2.shape[0]
+                dict_collector["three_cell"] += batch.x_3.shape[0]
+
+            elif cfg.model.model_domain == "cell":
+                dict_collector["zero_cell"] += batch.x_0.shape[0]
+                dict_collector["one_cell"] += batch.x_1.shape[0]
+                dict_collector["two_cell"] += batch.x_2.shape[0]
+                cell_sizes, cell_counts = torch.unique(batch.incidence_2.to_dense().sum(0), return_counts=True)
+
+
+    # Get current working dir
+    filename = f"{cfg.paths['root_dir']}/tables/dataset_statistics.csv"
+
+    dict_collector['dataset'] = cfg.dataset.parameters.data_name
+    dict_collector['domain'] = cfg.model.model_domain
+
+    df = pd.DataFrame.from_dict(dict_collector, orient='index')
+    if not os.path.exists(filename) == True:
+        # Save to csv file such as methods .... is a header
+        df.T.to_csv(filename, header=True)
+    else:
+        # read csv file with deader
+        df_saved = pd.read_csv(filename, index_col=0)
+        # add new row
+        df_saved = df_saved._append(dict_collector, ignore_index=True)
+        # write to csv file
+        df_saved.to_csv(filename)
+
+    # if cfg.model.model_domain == "cell":   
+    #     filename = f"{cfg.paths['root_dir']}/tables/cell_statistics.csv"
+    #     # Create a dict from two arrays
+    #     cell_dict = dict(zip(cell_sizes.long().tolist(), cell_counts.long().tolist()))
+
+    #     # Check if there are cells size of which greater than 10
+    #     n_large_cells = 0
+    #     subset_keys = [key for key in sorted(cell_dict.keys()) if key > 10]
+
+    #     for key in subset_keys:
+    #         n_large_cells += cell_dict.pop(key)
+
+    #     cell_dict["greater_than_10"] = n_large_cells
+
+    #     cell_dict['dataset'] = cfg.dataset.parameters.data_name
+    #     cell_dict['domain'] = cfg.model.model_domain
+
+    #     df = pd.DataFrame.from_dict(cell_dict, orient='index')
+    #     if not os.path.exists(filename) == True:
+    #         # Save to csv file such as methods .... is a header
+    #         df.T.to_csv(filename, header=True)
+    #     else:
+    #         # read csv file with deader
+    #         df_saved = pd.read_csv(filename, index_col=0)
+    #         # add new row
+    #         df_saved = df_saved._append(df.T, ignore_index=True)
+    #         # write to csv file
+    #         df_saved.to_csv(filename)
+
+
+    return 
+
+
+
+@hydra.main(
+    version_base="1.3", config_path="../configs", config_name="train.yaml"
+)
+def main(cfg: DictConfig) -> float | None:
+    """Main entry point for training.
+
+    :param cfg: DictConfig configuration composed by Hydra.
+    :return: Optional[float] with optimized metric value.
+    """
+    # apply extra utilities
+    # (e.g. ask for tags if none are provided in cfg, print cfg tree, etc.)
+    extras(cfg)
+
+
+    train(cfg)
+
+
+    # return optimized metric
+    return 
+
+
+if __name__ == "__main__":
+    main()

topobenchmarkx/stat.sh

@@ -0,0 +1,61 @@
+
+# Description: Main experiment script for GCN model.
+# ----Node regression datasets: US County Demographics----
+models=( 'simplicial/scn' 'cell/cwn' 'hypergraph/unignn2' )
+for model in ${models[*]}
+do
+
+
+python dataset_statistics.py \
+    dataset=us_country_demos \
+    model=$model \
+
+
+# ----Cocitation datasets----
+datasets=( 'cocitation_cora' 'cocitation_citeseer' 'cocitation_pubmed' )
+
+for dataset in ${datasets[*]}
+do
+  python dataset_statistics.py \
+    dataset=$dataset \
+    model=$model
+
+done
+
+# ----Graph regression dataset----
+# Train on ZINC dataset
+python dataset_statistics.py \
+  dataset=ZINC \
+  model=$model \
+  dataset.transforms.one_hot_node_degree_features.degrees_fields=x
+
+
+# ----Heterophilic datasets----
+
+datasets=( roman_empire amazon_ratings tolokers questions minesweeper )
+
+for dataset in ${datasets[*]}
+do
+  python dataset_statistics.py \
+    dataset=$dataset \
+    model=$model
+done
+
+# ----TU graph datasets----
+# MUTAG have very few samples, so we use a smaller batch size
+# Train on MUTAG dataset
+python dataset_statistics.py \
+  dataset=MUTAG \
+  model=$model
+
+# Train rest of the TU graph datasets
+datasets=( 'PROTEINS_TU' 'NCI1' 'NCI109' 'REDDIT-BINARY' 'IMDB-BINARY' 'IMDB-MULTI') # 
+
+for dataset in ${datasets[*]}
+do
+  python dataset_statistics.py \
+    dataset=$dataset \
+    model=$model
+done
+
+done