Qiming Sun

If you just need some CC code with GPU for matrix multiplication, I'd recommend copy the code from pyscf.cc module, replacing lib.einsum with cupy.einsum (plus some conversion from numpy array...

Please use `gpu4pyscf.dft.RKS(mol)` instead of `mol.RKS()`. The later creates a RKS object implemented in CPU. CPU module and GPU module are not compatible in this case. There needs a check...

I cannot reproduce the issue. I guess it is not a problem of gpu4pyscf. It's probably just a common issue for PES jobs. There are two tricks may be helpful...

The fluctuation is caused by the discontinuity in the weights of DFT grids wrt geometry. If you use a very small step size in scan, it probably will produce a...

There may be multiple libcusolver.so libraries on the system. I suspect an incorrect version is loaded. Could you run ``` LD_DEBUG=libs python -c 'import ctypes; ctypes.CDLL("libcusolver.so")' ``` to check which...

I'll close this issue for now since the problem cannot be reproduced. Please feel free to reopen it if it occurs again.

This function is not supposed to be used for an arbitrary Mole object. As the name suggests, it can only be called with the "sorted_mol" which is generated by the...

The precompiled pip wheels only include ptxas code for sm-70, 80, and 90. You might need to compile additional sm archs for your system from source code.

Yes, the default `auxbasis_response` is the issue. It will be changed to 2 in the future.

In gpu4pyscf, `jk.get_jk` only supports building J, K matrices for zeroth-order ERIs. Features like higher-order derivatives and arbitrary contraction patterns over multiple `mol` objects in PySCF will not be implemented...