Merge remote-tracking branch 'origin/main' into publication

Author: Azure Pipelines

.github/workflows/ci_docs.yml

@@ -90,7 +90,7 @@ jobs:
           done <<< $(cat changed-folders.txt)
           tree changed-notebooks
 
-      - uses: actions/upload-artifact@v3
+      - uses: actions/upload-artifact@v4
         if: ${{ matrix.target == 'html' && env.NB_DIRS != 0 }}
         with:
           name: notebooks-${{ github.sha }}
@@ -102,7 +102,7 @@ jobs:
 
       - name: Upload built docs
         if: ${{ matrix.target == 'html' }}
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
           name: docs-html-${{ github.sha }}
           path: _docs/build/html/

.notebooks/README.md

@@ -4,6 +4,7 @@ shadow folder for generated notebooks, no uploading here
 
 ```bash
 git checkout -b publication main
+export MAKEFLAGS="-j$(nproc)"
 make ipynb
 git commit -m "regenerate all notebooks"
 git push

README.md

@@ -97,3 +97,19 @@ pytest -v notebook.ipynb  --nbval
 # 3) generating notebooks outputs
 papermill in-notebook.ipynb out-notebook.ipynb
 ```
+
+## Development tips
+
+### Local Docs build
+
+You may want to build the documentation local without need to excrete all notebooks.
+In such case you can convert all scripts to ipython notebooks as dry run...
+
+```bash
+# set skip notebooks execution, just conversion
+export DRY_RUN=1
+# generate notebooks from scripts
+make ipynb
+# build the documentation
+make docs
+```

_requirements/devel.txt

@@ -3,7 +3,7 @@ virtualenv>=20.10
 # pytest>=7.0, <9.0
 # testing with own fork with extended cell timeout
 # https://github.com/Borda/nbval/archive/refs/heads/timeout-limit.zip
-ipython[notebook]>=8.0.0, <8.17.0
+ipython[notebook]>=8.0.0, <8.24.0
 urllib3  # for ipython
 jupytext>=1.10, <1.15  # converting
 papermill>=2.3.4, <2.5.0  # rendering

course_UvA-DL/01-introduction-to-pytorch/requirements.txt

@@ -1,2 +0,0 @@
-matplotlib
-lightning>=2.0.0

course_UvA-DL/02-activation-functions/requirements.txt

@@ -1,4 +1,2 @@
 torchvision
-matplotlib
 seaborn
-lightning>=2.0.0

course_UvA-DL/03-initialization-and-optimization/notebook.py

@@ -13,12 +13,12 @@
 import urllib.request
 from urllib.error import HTTPError
 
-import lightning as L
 import matplotlib.pyplot as plt
 
 # %matplotlib inline
 import matplotlib_inline.backend_inline
 import numpy as np
+import pytorch_lightning as pl
 import seaborn as sns
 import torch
 import torch.nn as nn
@@ -33,7 +33,7 @@
 sns.set()
 
 # %% [markdown]
-# Instead of the `set_seed` function as in Tutorial 3, we can use Lightning's built-in function `L.seed_everything`.
+# Instead of the `set_seed` function as in Tutorial 3, we can use Lightning's built-in function `pl.seed_everything`.
 # We will reuse the path variables `DATASET_PATH` and `CHECKPOINT_PATH` as in Tutorial 3.
 # Adjust the paths if necessary.
 
@@ -44,7 +44,7 @@
 CHECKPOINT_PATH = os.environ.get("PATH_CHECKPOINT", "saved_models/InitOptim/")
 
 # Seed everything
-L.seed_everything(42)
+pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
 torch.backends.cudnn.deterministic = True

course_UvA-DL/03-initialization-and-optimization/requirements.txt

@@ -1,4 +1,2 @@
 torchvision
-matplotlib
 seaborn
-lightning>=2.0.0

course_UvA-DL/04-inception-resnet-densenet/notebook.py

@@ -8,11 +8,11 @@
 from types import SimpleNamespace
 from urllib.error import HTTPError
 
-import lightning as L
 import matplotlib
 import matplotlib.pyplot as plt
 import matplotlib_inline.backend_inline
 import numpy as np
+import pytorch_lightning as pl
 import seaborn as sns
 import tabulate
 import torch
@@ -23,7 +23,6 @@
 
 # %matplotlib inline
 from IPython.display import HTML, display
-from lightning.pytorch.callbacks import LearningRateMonitor, ModelCheckpoint
 from PIL import Image
 from torchvision import transforms
 from torchvision.datasets import CIFAR10
@@ -47,7 +46,7 @@
 
 
 # Function for setting the seed
-L.seed_everything(42)
+pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
 torch.backends.cudnn.deterministic = True
@@ -137,9 +136,9 @@
 # We need to do a little trick because the validation set should not use the augmentation.
 train_dataset = CIFAR10(root=DATASET_PATH, train=True, transform=train_transform, download=True)
 val_dataset = CIFAR10(root=DATASET_PATH, train=True, transform=test_transform, download=True)
-L.seed_everything(42)
+pl.seed_everything(42)
 train_set, _ = torch.utils.data.random_split(train_dataset, [45000, 5000])
-L.seed_everything(42)
+pl.seed_everything(42)
 _, val_set = torch.utils.data.random_split(val_dataset, [45000, 5000])
 
 # Loading the test set
@@ -193,12 +192,12 @@
 
 # %%
 # Setting the seed
-L.seed_everything(42)
+pl.seed_everything(42)
 
 # %% [markdown]
 # Thus, in the future, we don't have to define our own `set_seed` function anymore.
 #
-# In PyTorch Lightning, we define `L.LightningModule`'s (inheriting from `Module`) that organize our code into 5 main sections:
+# In PyTorch Lightning, we define `pl.LightningModule`'s (inheriting from `Module`) that organize our code into 5 main sections:
 #
 # 1. Initialization (`__init__`), where we create all necessary parameters/models
 # 2. Optimizers (`configure_optimizers`) where we create the optimizers, learning rate scheduler, etc.
@@ -215,7 +214,7 @@
 
 
 # %%
-class CIFARModule(L.LightningModule):
+class CIFARModule(pl.LightningModule):
     def __init__(self, model_name, model_hparams, optimizer_name, optimizer_hparams):
         """CIFARModule.
 
@@ -296,6 +295,7 @@ def test_step(self, batch, batch_idx):
 
 # %%
 # Callbacks
+from pytorch_lightning.callbacks import LearningRateMonitor, ModelCheckpoint  # noqa: E402
 
 # %% [markdown]
 # To allow running multiple different models with the same Lightning module, we define a function below that maps a model name to the model class.
@@ -350,7 +350,7 @@ def train_model(model_name, save_name=None, **kwargs):
         save_name = model_name
 
     # Create a PyTorch Lightning trainer with the generation callback
-    trainer = L.Trainer(
+    trainer = pl.Trainer(
         default_root_dir=os.path.join(CHECKPOINT_PATH, save_name),  # Where to save models
         # We run on a single GPU (if possible)
         accelerator="auto",
@@ -374,7 +374,7 @@ def train_model(model_name, save_name=None, **kwargs):
         # Automatically loads the model with the saved hyperparameters
         model = CIFARModule.load_from_checkpoint(pretrained_filename)
     else:
-        L.seed_everything(42)  # To be reproducible
+        pl.seed_everything(42)  # To be reproducible
         model = CIFARModule(model_name=model_name, **kwargs)
         trainer.fit(model, train_loader, val_loader)
         model = CIFARModule.load_from_checkpoint(

course_UvA-DL/04-inception-resnet-densenet/requirements.txt

@@ -1,6 +1,4 @@
 torchvision
-matplotlib
 seaborn
 tabulate
-lightning>=2.0.0
 tensorboard

course_UvA-DL/05-transformers-and-MH-attention/MHAttention.py

@@ -22,14 +22,14 @@
 from functools import partial
 from urllib.error import HTTPError
 
-# PyTorch Lightning
-import lightning as L
-
 # Plotting
 import matplotlib
 import matplotlib.pyplot as plt
 import matplotlib_inline.backend_inline
 import numpy as np
+
+# PyTorch Lightning
+import pytorch_lightning as pl
 import seaborn as sns
 
 # PyTorch
@@ -41,7 +41,7 @@
 
 # Torchvision
 import torchvision
-from lightning.pytorch.callbacks import ModelCheckpoint
+from pytorch_lightning.callbacks import ModelCheckpoint
 from torchvision import transforms
 from torchvision.datasets import CIFAR100
 from tqdm.notebook import tqdm
@@ -58,7 +58,7 @@
 CHECKPOINT_PATH = os.environ.get("PATH_CHECKPOINT", "saved_models/Transformers/")
 
 # Setting the seed
-L.seed_everything(42)
+pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
 torch.backends.cudnn.deterministic = True
@@ -246,7 +246,7 @@ def scaled_dot_product(q, k, v, mask=None):
 
 # %%
 seq_len, d_k = 3, 2
-L.seed_everything(42)
+pl.seed_everything(42)
 q = torch.randn(seq_len, d_k)
 k = torch.randn(seq_len, d_k)
 v = torch.randn(seq_len, d_k)
@@ -748,7 +748,7 @@ def get_lr_factor(self, epoch):
 
 
 # %%
-class TransformerPredictor(L.LightningModule):
+class TransformerPredictor(pl.LightningModule):
     def __init__(
         self,
         input_dim,
@@ -965,7 +965,7 @@ def test_step(self, batch, batch_idx):
 
 # %% [markdown]
 # Finally, we can create a training function similar to the one we have seen in Tutorial 5 for PyTorch Lightning.
-# We create a `L.Trainer` object, running for $N$ epochs, logging in TensorBoard, and saving our best model based on the validation.
+# We create a `pl.Trainer` object, running for $N$ epochs, logging in TensorBoard, and saving our best model based on the validation.
 # Afterward, we test our models on the test set.
 # An additional parameter we pass to the trainer here is `gradient_clip_val`.
 # This clips the norm of the gradients for all parameters before taking an optimizer step and prevents the model
@@ -983,7 +983,7 @@ def train_reverse(**kwargs):
     # Create a PyTorch Lightning trainer with the generation callback
     root_dir = os.path.join(CHECKPOINT_PATH, "ReverseTask")
     os.makedirs(root_dir, exist_ok=True)
-    trainer = L.Trainer(
+    trainer = pl.Trainer(
         default_root_dir=root_dir,
         callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
         accelerator="auto",
@@ -1444,7 +1444,7 @@ def train_anomaly(**kwargs):
     # Create a PyTorch Lightning trainer with the generation callback
     root_dir = os.path.join(CHECKPOINT_PATH, "SetAnomalyTask")
     os.makedirs(root_dir, exist_ok=True)
-    trainer = L.Trainer(
+    trainer = pl.Trainer(
         default_root_dir=root_dir,
         callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
         accelerator="auto",

course_UvA-DL/05-transformers-and-MH-attention/requirements.txt

@@ -1,4 +1,2 @@
 torchvision
-matplotlib
 seaborn
-lightning>=2.0.0

course_UvA-DL/06-graph-neural-networks/overview.py

@@ -11,7 +11,7 @@
 from urllib.error import HTTPError
 
 # PyTorch Lightning
-import lightning as L
+import pytorch_lightning as pl
 
 # PyTorch
 import torch
@@ -25,7 +25,7 @@
 import torch_geometric.nn as geom_nn
 
 # PL callbacks
-from lightning.pytorch.callbacks import ModelCheckpoint
+from pytorch_lightning.callbacks import ModelCheckpoint
 from torch import Tensor
 
 AVAIL_GPUS = min(1, torch.cuda.device_count())
@@ -36,7 +36,7 @@
 CHECKPOINT_PATH = os.environ.get("PATH_CHECKPOINT", "saved_models/GNNs/")
 
 # Setting the seed
-L.seed_everything(42)
+pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
 torch.backends.cudnn.deterministic = True
@@ -592,7 +592,7 @@ def forward(self, x, *args, **kwargs):
 
 
 # %%
-class NodeLevelGNN(L.LightningModule):
+class NodeLevelGNN(pl.LightningModule):
     def __init__(self, model_name, **model_kwargs):
         super().__init__()
         # Saving hyperparameters
@@ -654,13 +654,13 @@ def test_step(self, batch, batch_idx):
 
 # %%
 def train_node_classifier(model_name, dataset, **model_kwargs):
-    L.seed_everything(42)
+    pl.seed_everything(42)
     node_data_loader = geom_data.DataLoader(dataset, batch_size=1)
 
     # Create a PyTorch Lightning trainer
     root_dir = os.path.join(CHECKPOINT_PATH, "NodeLevel" + model_name)
     os.makedirs(root_dir, exist_ok=True)
-    trainer = L.Trainer(
+    trainer = pl.Trainer(
         default_root_dir=root_dir,
         callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
         accelerator="auto",
@@ -676,7 +676,7 @@ def train_node_classifier(model_name, dataset, **model_kwargs):
         print("Found pretrained model, loading...")
         model = NodeLevelGNN.load_from_checkpoint(pretrained_filename)
     else:
-        L.seed_everything()
+        pl.seed_everything()
         model = NodeLevelGNN(
             model_name=model_name, c_in=dataset.num_node_features, c_out=dataset.num_classes, **model_kwargs
         )
@@ -892,7 +892,7 @@ def forward(self, x, edge_index, batch_idx):
 
 
 # %%
-class GraphLevelGNN(L.LightningModule):
+class GraphLevelGNN(pl.LightningModule):
     def __init__(self, **model_kwargs):
         super().__init__()
         # Saving hyperparameters
@@ -941,12 +941,12 @@ def test_step(self, batch, batch_idx):
 
 # %%
 def train_graph_classifier(model_name, **model_kwargs):
-    L.seed_everything(42)
+    pl.seed_everything(42)
 
     # Create a PyTorch Lightning trainer with the generation callback
     root_dir = os.path.join(CHECKPOINT_PATH, "GraphLevel" + model_name)
     os.makedirs(root_dir, exist_ok=True)
-    trainer = L.Trainer(
+    trainer = pl.Trainer(
         default_root_dir=root_dir,
         callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
         accelerator="cuda",
@@ -962,7 +962,7 @@ def train_graph_classifier(model_name, **model_kwargs):
         print("Found pretrained model, loading...")
         model = GraphLevelGNN.load_from_checkpoint(pretrained_filename)
     else:
-        L.seed_everything(42)
+        pl.seed_everything(42)
         model = GraphLevelGNN(
             c_in=tu_dataset.num_node_features,
             c_out=1 if tu_dataset.num_classes == 2 else tu_dataset.num_classes,

course_UvA-DL/06-graph-neural-networks/requirements.txt

@@ -3,4 +3,3 @@ torch-sparse
 torch-cluster
 torch-spline-conv
 torch-geometric>=2.0.0,<2.5.0
-lightning>=2.0.0

course_UvA-DL/07-deep-energy-based-generative-models/requirements.txt

@@ -1,4 +1,3 @@
 torchvision
 matplotlib
 tensorboard
-pytorch-lightning>=2.0.0