Add PyG dataset classes (#71)

* Add dataloader classes * [Bug] rogue unused argument * Add multiprocessing support * Fix bare excepts * Remove bare excepts * [Logging] update try except imports to use logging * isort * lint * [Dataloader] add base dataloader class * [Dataloader] add len * torch geometric dataset * add init * remove old dataloaders * refactor mp constructor * add rename arg to AF2 download and fix docstring * add docstrings * add new dataset classes to namespace * add transforms and docstrings * add tests for dataloaders * add tensor equality util * update docs * add dataloader tutorial * fix dataloader tests * update testing utils to have optional import for torch equality checks * fix bare except * remove torch.testing as it breaks support with previous versions <1.11 * black, isort * add torch geometric install * remove pyg from matrix tests * add skip if pyg not installed * typo * add option pyg imports * set imports to optional * make torch imports optional in tests * use pathlib instead of string paths * use tuple of exception classes * wrap dataset imports in try/except * fix to obsolete PDB download * Fix tests * fix chain graph return * add pyg install * rm pyg, ignore dataloader test * ignore dataloader tutorial * fix renaming errors in AF2 download
a-r-j · Apr 4, 2022 · d0f872b · d0f872b
1 parent a958b28
commit d0f872b
Show file tree

Hide file tree

Showing 20 changed files with 1,576 additions and 41 deletions.
diff --git a/.github/workflows/build.yaml b/.github/workflows/build.yaml
@@ -67,4 +67,4 @@ jobs:
       - name: Run unit tests and generate coverage report
         run: pytest .
       - name: Test notebook execution
-        run: pytest --nbval-lax notebooks/ --current-env
+        run: pytest --nbval-lax notebooks/ --current-env --ignore-glob="notebooks/dataloader_tutorial.ipynb"
diff --git a/docs/source/datasets.rst b/docs/source/datasets.rst
@@ -10,6 +10,7 @@ Summaries
    :maxdepth: 2
    :glob:
 
+   notebooks/dataloader_tutorial.nblink
    datasets/pscdb
    notebooks/pscdb_processing.nblink
    notebooks/pscdb_baselines.nblink

diff --git a/docs/source/notebooks/dataloader_tutorial.nblink b/docs/source/notebooks/dataloader_tutorial.nblink
@@ -0,0 +1,3 @@
+{
+    "path": "../../../notebooks/dataloader_tutorial.ipynb"
+}
diff --git a/graphein/grn/features/node_features.py b/graphein/grn/features/node_features.py
@@ -1,14 +1,19 @@
+import logging
+
 from graphein.utils.utils import import_message
 
+log = logging.getLogger(__name__)
+
 try:
     from bioservices import HGNC, UniProt
 except ImportError:
-    import_message(
+    message = import_message(
         submodule="graphein.grn.features.node_features",
         package="bioservices",
         conda_channel="bioconda",
         pip_install=True,
     )
+    log.warning(message)
 
 
 def add_sequence_to_nodes(n, d):

diff --git a/graphein/ml/__init__.py b/graphein/ml/__init__.py
@@ -0,0 +1,10 @@
+from .conversion import GraphFormatConvertor
+
+try:
+    from .datasets import (
+        InMemoryProteinGraphDataset,
+        ProteinGraphDataset,
+        ProteinGraphListDataset,
+    )
+except (ImportError, NameError):
+    pass
diff --git a/graphein/ml/datasets/__init__.py b/graphein/ml/datasets/__init__.py
@@ -0,0 +1,5 @@
+from .torch_geometric_dataset import (
+    InMemoryProteinGraphDataset,
+    ProteinGraphDataset,
+    ProteinGraphListDataset,
+)