You can find here example notebooks provided by the AIHWKIT team.
You can open any page of the documentation as a notebook in colab (there is a button directly on said pages) but they are also listed here if you need to:
Notebook | Description | |
---|---|---|
Analog training of LeNet5 on ReRam analog device | Training the LeNet5 neural network with MNIST dataset and the Analog SGD optimizer simulated on the analog resistive random-access memory with soft bounds (ReRam) device | |
Analog Training of LeNet5 with the Tiki Taka optimizer and ReRAM analog device | Training the LeNet5 neural network with Tiki Taka analog optimizer on MNIST dataset, simulated on the the analog resistive random-access memory with soft bounds (ReRam) device | |
Analog Training of LeNet5 with hardware aware training for inference on a PCM device | Training the LeNet5 neural network with hardware aware training using the Inference RPU Config to optimize inference on a PCM device. | |
Experiments to study the sensitivity of neural networks to non-idealities of Crossbar implementations | Studying the sensitivity of a LeNet5 neural network to synaptic devices and peripheral circuit non-idealities. | |
IBM Analog Fusion chip conversion utility example | Provide examples of how to convert the weights of a neural network model to a list of conductance values that can be programmed in the Fusion chip. The notebook also shows the errors between the orginal model and the new model after programming the analog chip. |