📄️ TensorFlow and Keras
Deep learning model development with TensorFlow and Keras, from the first layer to a served model. Duration: 9h, with applied projects, a 40-question exam and a verifiable certificate.
📄️ 1. Tensors and graphs
Module 1 of the TensorFlow and Keras premium course: the difference between a tensor and a variable, eager versus graph execution, and what tf.function actually does.
📄️ 2. Sequential API
Module 2 of the TensorFlow and Keras premium course: build, compile and train a model with Sequential, read its summary and know where it stops being enough.
📄️ 3. Functional API
Module 3 of the TensorFlow and Keras premium course: build non-linear architectures, merge branches, share weights and weight several losses.
📄️ 4. Custom layers
Module 4 of the TensorFlow and Keras premium course: write a layer with build and call, handle training mode, serialise with get_config and override train_step.
📄️ 5. tf.data pipelines
Module 5 of the TensorFlow and Keras premium course: compose an efficient tf.data pipeline, understand why operation order matters and remove accelerator starvation.
📄️ 6. Callbacks
Module 6 of the TensorFlow and Keras premium course: ModelCheckpoint, EarlyStopping, ReduceLROnPlateau and custom callbacks, with the settings that actually matter.
📄️ 7. TensorBoard
Module 7 of the TensorFlow and Keras premium course: instrument a training run, compare experiments, read weight histograms and diagnose with the profiler.
📄️ 8. Transfer learning
Module 8 of the TensorFlow and Keras premium course: reuse a pretrained network, freeze then unfreeze layers, and avoid the BatchNormalization trap.
📄️ 9. Distributed training
Module 9 of the TensorFlow and Keras premium course: MirroredStrategy, global batch size, learning rate scaling and mixed precision.
📄️ 10. SavedModel and serving
Module 10 of the TensorFlow and Keras premium course: export as a SavedModel, inspect signatures, serve the model over HTTP and manage versions.
📄️ Recap and exam
Complete recap of the TensorFlow and Keras premium course: tensors, sequential and functional APIs, tf.data, callbacks, transfer learning, distribution and serving, then the 40-question exam.