Paper link: https://arxiv.org/abs/2306.14962
The code backprop_scaling_numerics.py allows one to reproduce the results of Figure 6 and Figure 9 in the article.
- Pennylane
- Jax/Jaxlib
- Optax
- Matplotlib
You can simply install the dependencies from the requirements.txt file via pip install -r requirements.txt.
Contact: [email protected]