Feature/multi gpu

.gitignore

@@ -112,4 +112,7 @@ venv.bak/
 
 # direnv configuration file
 .envrc
+
+# PowerPoint checkpoint files
 /~$machine_learning_with_python.pptx
+/~$multi_gpu_machine_learning.pptx

README.md

@@ -9,17 +9,15 @@ For information on the training, see the website
 
 1. [`machine_learning_with_python.pptx`](machine_learning_with_python.pptx): PowerPoint
    presentation used for the training.
+1. ['multi_gpu_machine_learning.pptx'](multi_gpu_machine_learning.pptx): PowerPoint presentation used for the
+   multi-GPU session of the training.
 1. [`hands-on`](hands-on): Jupyter notebooks for hands-on sessions.
 1. [`source-code`](source-code): sample code written to develop the slides and
    illustrate concepts.
 1. [`environment.yml`](environment.yml): CPU-only conda environment for the
    core teaching material, intended to be cross-platform.
 1. [`environment_gpu.yml`](environment_gpu.yml): GPU-enabled conda environment
    for the instructor laptop, using its validated CUDA configuration.
-1. [`environment_keras.yml`](environment_keras.yml): CPU-only Linux conda
-   environment for the legacy Keras examples.
-1. [`environment_keras_gpu.yml`](environment_keras_gpu.yml): Linux/NVIDIA conda
-   environment for running the legacy Keras examples with GPU acceleration.
 1. [`machine_learning_with_python_linux64_conda_specs.txt`](machine_learning_with_python_linux64_conda_specs.txt):
    conda environment specification file specific for 64-bit Linux to precisely
    reproduce the environment on which the code was developed.

environment.yml

@@ -17,10 +17,10 @@ dependencies:
 
   # Model interpretation and parameter-optimization examples.
   - lime
-  - hyperopt
-  - bayesian-optimization
+  - optuna
 
   # Notebook and visualization stack used by the core teaching material.
   - jupyterlab
+  - ipywidgets
   - matplotlib
   - bokeh

environment_gpu.yml

@@ -10,14 +10,18 @@ dependencies:
   - torchvision=0.28
   - cuda-version=12.9
 
-  # Keep the teaching stack aligned with the CPU environment.
+  # Core numerical and machine-learning stack.
   - numpy
   - scipy
   - pandas
   - scikit-learn
+
+  # Model interpretation and parameter-optimization examples.
   - lime
-  - hyperopt
-  - bayesian-optimization
+  - optuna
+
+  # Notebook and visualization stack used by the core teaching material.
   - jupyterlab
+  - ipywidgets
   - matplotlib
   - bokeh

source-code/README.md

@@ -9,9 +9,8 @@ to create it.  There is some material not covered in the presentation as well.
 * [`pytorch`](pytorch): illustration of using PyTorch for machine learning.
 * [`pytorch-lightning/`](pytorch-lightning/): illustration of using PyTorch
   Lightning  for machine learning.
-* [`parameter-optimization`](parameter-optimization): example of parameter
-  optimization kusing hyperopt, although the examples do not optimize
-  hyperparameters in machine learning, that would be very similar.
+* [`parameter-optimization`](parameter-optimization): methodology-first Optuna
+  examples, from a minimal study to a realistic PINN case study.
 * [`scikit-learn`](scikit-learn): examples of scikit-learn for machine
   learning, examples are provided forsupervised (regression and classification)
   and unsupervised (clustering) learnign, as well as dimensionality reduction

source-code/keras/README.md

@@ -15,6 +15,10 @@ machine learning tasks.
 1. `mnist_cnn.pbs`: example PBS script to exectute the `mnist_cnn.ipynb`
     notebook in batch as a job.
 1. `Flatland`: machine learning adventures in Flatland.
+1. [`environment_keras.yml`](environment_keras.yml): CPU-only Linux conda
+   environment for the legacy Keras examples.
+1. [`environment_keras_gpu.yml`](environment_keras_gpu.yml): Linux/NVIDIA conda
+   environment for running the legacy Keras examples with GPU acceleration.
 
 ## References
 Excellent [tutorial](https://github.com/leriomaggio/deep-learning-keras-tensorflow).