dnc_pytorch/README.md

Differentiable Neural Computer in PyTorch

This is PyTorch port of DeepMind's Differentiable Neural Computer (DNC).

The original code was written with TensorFlow v1, which is not straightforward to set up in modern tech stack, such as Apple Scilicon and Python 3.8+.

This code requires only PyTorch >= 1.10.

The code structure and interfaces are kept almost same.

You can run repeat-copy task with python train.py.