TensorFlow Taskrunner Workspace with Keras 3

[tanwarsh]

TensorFlow Taskrunner Workspace with Keras 3, showcasing custom model and training and testing loops.

[rahulga1]

can you highlight which part is tensorflow specific in this workspace.

[tanwarsh]

can you highlight which part is tensorflow specific in this workspace.

CNNModel in taskruner.py implements a custom model using TensorFlow with test_step and train_step functions for evaluation and training, respectively.

[teoparvanov]

@tanwarsh , what does this workspace bring in addition to the keras/mnist one? Isn't this showcasing how to use TensorFlow as a Keras 3 backend?

PS: couldn't we instead move the existing keras/mnist workspace under keras/tf/mnist?

[tanwarsh]

@tanwarsh , what does this workspace bring in addition to the keras/mnist one? Isn't this showcasing how to use TensorFlow as a Keras 3 backend?

PS: couldn't we instead move the existing keras/mnist workspace under keras/tf/mnist?

Thank you for the review @teoparvanov . The keras/mnist workspace uses Tensorflow internally but with this workspace we are showcasing how training and test loops can be written with Tensorflow. (train_step and test_step) same as JAX and pytorch workspace.