UvA DL Tutorials: Updating PL API to >1.8

course_UvA-DL/01-introduction-to-pytorch/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 1: Introduction to PyTorch"
 author: Phillip Lippe
 created: 2021-08-27
-updated: 2021-11-29
+updated: 2023-01-04
 license: CC BY-SA
 description: |
  This tutorial will give a short introduction to PyTorch basics, and get you setup for writing your own neural networks.

course_UvA-DL/01-introduction-to-pytorch/Introduction_to_PyTorch.py

@@ -455,7 +455,7 @@
 
 # Additionally, some operations on a GPU are implemented stochastic for efficiency
 # We want to ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 # %% [markdown]

course_UvA-DL/02-activation-functions/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 2: Activation Functions"
 author: Phillip Lippe
 created: 2021-08-27
-updated: 2021-08-27
+updated: 2023-01-04
 license: CC BY-SA
 description: |
  In this tutorial, we will take a closer look at (popular) activation functions and investigate their effect on optimization properties in neural networks.

course_UvA-DL/02-activation-functions/Activation_Functions.py

@@ -64,7 +64,7 @@ def set_seed(seed):
 
 # Additionally, some operations on a GPU are implemented stochastic for efficiency
 # We want to ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 # Fetching the device that will be used throughout this notebook

course_UvA-DL/03-initialization-and-optimization/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 3: Initialization and Optimization"
 author: Phillip Lippe
 created: 2021-08-27
-updated: 2021-11-29
+updated: 2023-01-04
 license: CC BY-SA
 tags:
  - Image

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

@@ -47,7 +47,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 # Fetching the device that will be used throughout this notebook
@@ -937,8 +937,8 @@ def pathological_curve_loss(w1, w2):
 def plot_curve(
  curve_fn, x_range=(-5, 5), y_range=(-5, 5), plot_3d=False, cmap=cm.viridis, title="Pathological curvature"
 ):
- fig = plt.figure()
- ax = fig.gca(projection="3d") if plot_3d else fig.gca()
+ _ = plt.figure()
+ ax = plt.axes(projection="3d") if plot_3d else plt.axes()
 
  x = torch.arange(x_range[0], x_range[1], (x_range[1] - x_range[0]) / 100.0)
  y = torch.arange(y_range[0], y_range[1], (y_range[1] - y_range[0]) / 100.0)

course_UvA-DL/04-inception-resnet-densenet/.meta.yaml

@@ -1,7 +1,7 @@
 title: "Tutorial 4: Inception, ResNet and DenseNet"
 author: Phillip Lippe
 created: 2021-08-27
-updated: 2021-11-29
+updated: 2023-01-04
 license: CC BY-SA
 tags:
  - Image
@@ -18,5 +18,6 @@ requirements:
  - matplotlib
  - seaborn
  - tabulate
+ - pytorch-lightning>=1.8
 accelerator:
  - GPU

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

@@ -49,7 +49,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
@@ -348,7 +348,8 @@ def train_model(model_name, save_name=None, **kwargs):
  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)
- gpus=1 if str(device) == "cuda:0" else 0,
+ accelerator="gpu" if str(device).startswith("cuda") else "cpu",
+ devices=1,
  # How many epochs to train for if no patience is set
  max_epochs=180,
  callbacks=[
@@ -357,7 +358,7 @@ def train_model(model_name, save_name=None, **kwargs):
  ), # Save the best checkpoint based on the maximum val_acc recorded. Saves only weights and not optimizer
  LearningRateMonitor("epoch"),
  ], # Log learning rate every epoch
- progress_bar_refresh_rate=1,
+ enable_progress_bar=True,
  ) # In case your notebook crashes due to the progress bar, consider increasing the refresh rate
  trainer.logger._log_graph = True # If True, we plot the computation graph in tensorboard
  trainer.logger._default_hp_metric = None # Optional logging argument that we don't need

course_UvA-DL/05-transformers-and-MH-attention/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 5: Transformers and Multi-Head Attention"
 author: Phillip Lippe
 created: 2021-06-30
-updated: 2021-11-29
+updated: 2023-01-04
 license: CC BY-SA
 build: 0
 tags:
@@ -19,5 +19,6 @@ requirements:
  - torchvision
  - matplotlib
  - seaborn
+ - pytorch-lightning>=1.8
 accelerator:
  - GPU

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

@@ -61,7 +61,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
@@ -979,10 +979,11 @@ def train_reverse(**kwargs):
  trainer = pl.Trainer(
  default_root_dir=root_dir,
  callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
- gpus=1 if str(device).startswith("cuda") else 0,
+ accelerator="gpu" if str(device).startswith("cuda") else "cpu",
+ devices=1,
  max_epochs=10,
  gradient_clip_val=5,
- progress_bar_refresh_rate=1,
+ enable_progress_bar=True,
  )
  trainer.logger._default_hp_metric = None # Optional logging argument that we don't need
 
@@ -1439,10 +1440,11 @@ def train_anomaly(**kwargs):
  trainer = pl.Trainer(
  default_root_dir=root_dir,
  callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
- gpus=1 if str(device).startswith("cuda") else 0,
+ accelerator="gpu" if str(device).startswith("cuda") else "cpu",
+ devices=1,
  max_epochs=100,
  gradient_clip_val=2,
- progress_bar_refresh_rate=1,
+ enable_progress_bar=True,
  )
  trainer.logger._default_hp_metric = None # Optional logging argument that we don't need
 

course_UvA-DL/06-graph-neural-networks/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 6: Basics of Graph Neural Networks"
 author: Phillip Lippe
 created: 2021-06-07
-updated: 2021-12-04
+updated: 2023-01-04
 license: CC BY-SA
 build: 0
 tags:
@@ -19,10 +19,11 @@ description: |
  The full list of tutorials can be found at https://uvadlc-notebooks.rtfd.io.
 requirements:
  - torch-scatter
- - torch-sparse<0.6.13
+ - torch-sparse
  - torch-cluster
  - torch-spline-conv
- - torch-geometric==2.0.2
+ - torch-geometric
+ - pytorch-lightning>=1.8
 pip__find-link:
  # - https://pytorch-geometric.com/whl/torch-1.8.0+cu101.html
  - https://pytorch-geometric.com/whl/torch-%(TORCH_MAJOR_DOT_MINOR)s.0+%(DEVICE)s.html

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

@@ -39,7 +39,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 # %% [markdown]
@@ -634,7 +634,7 @@ def test_step(self, batch, batch_idx):
 # Additionally to the Lightning module, we define a training function below.
 # As we have a single graph, we use a batch size of 1 for the data loader and share the same data loader for the train,
 # validation, and test set (the mask is picked inside the Lightning module).
-# Besides, we set the argument `progress_bar_refresh_rate` to zero as it usually shows the progress per epoch,
+# Besides, we set the argument `enable_progress_bar` to False as it usually shows the progress per epoch,
 # but an epoch only consists of a single step.
 # If you have downloaded the pre-trained models in the beginning of the tutorial, we load those instead of training from scratch.
 # Finally, we test the model and return the results.
@@ -651,9 +651,10 @@ def train_node_classifier(model_name, dataset, **model_kwargs):
  trainer = pl.Trainer(
  default_root_dir=root_dir,
  callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
- gpus=AVAIL_GPUS,
+ accelerator="gpu" if AVAIL_GPUS > 0 else "cpu",
+ devices=max(1, AVAIL_GPUS),
  max_epochs=200,
- progress_bar_refresh_rate=0,
+ enable_progress_bar=False,
  ) # 0 because epoch size is 1
  trainer.logger._default_hp_metric = None # Optional logging argument that we don't need
 
@@ -932,9 +933,10 @@ def train_graph_classifier(model_name, **model_kwargs):
  trainer = pl.Trainer(
  default_root_dir=root_dir,
  callbacks=[ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc")],
- gpus=AVAIL_GPUS,
+ accelerator="gpu" if AVAIL_GPUS > 0 else "cpu",
+ devices=max(1, AVAIL_GPUS),
  max_epochs=500,
- progress_bar_refresh_rate=0,
+ enable_progress_bar=False,
  )
  trainer.logger._default_hp_metric = None
 

course_UvA-DL/07-deep-energy-based-generative-models/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 7: Deep Energy-Based Generative Models"
 author: Phillip Lippe
 created: 2021-07-12
-updated: 2021-07-12
+updated: 2023-01-04
 license: CC BY-SA
 build: 0
 tags:
@@ -22,6 +22,7 @@ requirements:
  - torchvision
  - matplotlib
  - tensorboard
+ - pytorch-lightning>=1.8
 accelerator:
  - CPU
  - GPU

course_UvA-DL/07-deep-energy-based-generative-models/Deep_Energy_Models.py

@@ -44,7 +44,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
@@ -640,7 +640,8 @@ def train_model(**kwargs):
  # Create a PyTorch Lightning trainer with the generation callback
  trainer = pl.Trainer(
  default_root_dir=os.path.join(CHECKPOINT_PATH, "MNIST"),
- gpus=1 if str(device).startswith("cuda") else 0,
+ accelerator="gpu" if str(device).startswith("cuda") else "cpu",
+ devices=1,
  max_epochs=60,
  gradient_clip_val=0.1,
  callbacks=[
@@ -650,7 +651,7 @@ def train_model(**kwargs):
  OutlierCallback(),
  LearningRateMonitor("epoch"),
  ],
- progress_bar_refresh_rate=1,
+ enable_progress_bar=True,
  )
  # Check whether pretrained model exists. If yes, load it and skip training
  pretrained_filename = os.path.join(CHECKPOINT_PATH, "MNIST.ckpt")

course_UvA-DL/08-deep-autoencoders/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 8: Deep Autoencoders"
 author: Phillip Lippe
 created: 2021-07-12
-updated: 2021-07-12
+updated: 2023-01-04
 license: CC BY-SA
 build: 0
 tags:
@@ -22,6 +22,7 @@ requirements:
  - torchvision
  - matplotlib
  - seaborn
+ - pytorch-lightning>=1.8
 accelerator:
  - CPU
  - GPU

course_UvA-DL/08-deep-autoencoders/Deep_Autoencoders.py

@@ -41,7 +41,7 @@
 pl.seed_everything(42)
 
 # Ensure that all operations are deterministic on GPU (if used) for reproducibility
-torch.backends.cudnn.determinstic = True
+torch.backends.cudnn.deterministic = True
 torch.backends.cudnn.benchmark = False
 
 device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
@@ -385,7 +385,8 @@ def train_cifar(latent_dim):
  # Create a PyTorch Lightning trainer with the generation callback
  trainer = pl.Trainer(
  default_root_dir=os.path.join(CHECKPOINT_PATH, "cifar10_%i" % latent_dim),
- gpus=1 if str(device).startswith("cuda") else 0,
+ accelerator="gpu" if str(device).startswith("cuda") else "cpu",
+ devices=1,
  max_epochs=500,
  callbacks=[
  ModelCheckpoint(save_weights_only=True),

course_UvA-DL/09-normalizing-flows/.meta.yml

@@ -1,7 +1,7 @@
 title: "Tutorial 9: Normalizing Flows for Image Modeling"
 author: Phillip Lippe
 created: 2021-06-07
-updated: 2021-06-16
+updated: 2023-01-04
 license: CC BY-SA
 build: 0
 tags:
@@ -25,6 +25,7 @@ requirements:
  - matplotlib
  - seaborn
  - tabulate
+ - pytorch-lightning>=1.8
 accelerator:
  - CPU
  - GPU