AlexanderMath
AlexanderMath
I save model during training with G: 1.42529; D: 1.17999 After loading discriminator and generator I get G: -26.45726; D: 0.89195 It looks like discriminator behaves differently. Removing 'd_optimizer.step' gives...
Our DFT function considers only a single molecule. We could compute DFT for a batch of molecules using `jax.vmap`. **Problem:** This consumes more on-chip memory. **Idea:** Take a molecule `a`....
The gif below visualizes the numerical distribution of all tensors during our DFT computation. TODO: 1. Refactor code for plot into `utils.py` 2. Try to loss scaling, operation pre-scaling/pre-shifting, ......
Transform the Jax graph to perform everything in float64 except a set of user-specified operations. May not be possible, we need to think about what that would look like as...
#55 outlines several redundancies in `ERI`. We want to determine which integrals are above a threshold (to be computed) on the IPU. Tasks 1. write python for loop code which...
Our current `ipu_eigh` uses the Jacobi algorithm. It is believed for other hw accelerators that the QR algorithm becomes faster than jacobi for larger `d>=512`. Since we are targeting `d>=512`...
Port all of libcint to poplar.
Remove `experimental_nanoDFT` with `nanoDFT.py`
List of similarities between DFT and Deep Learning 1. Both solve optimization problem 2. Both require initialization techniques (e.g. minao init or xavier init). 3. Both utilize momentum techniques (e.g....