ExtremelyAsyn -restart tested #88
Conversation
joannaqw
commented
Apr 25, 2024
joannaqw
commented
- Made a new class extremeAsynLASSCF, which is a subclass of LASSCFNoSymm
- within the new class, modified rdm_cycle and ci_cycle to read RDMs, write Hamiltonians and break
- added reading and writing orbitals (mo_coeff) within the kernel function
|
Could you resolve conflicts please? (Pull master to your local branch, resolve the conflict as you see fit, commit, and push to your remote branch and it will update here.) The reason that's important is that it will enable me to run the Github Actions unittest workflow to see if this breaks anything. |
| @@ -117,48 +118,60 @@ def get_init_guess_rdm (las, mo_coeff=None, h2eff_sub=None): | |||
| def _ddm (dm1frs0, dm1frs1): | |||
| return np.concatenate ([(d1 - d0).ravel () for d1, d0 in zip (dm1frs0, dm1frs1)]) | |||
|
|
|||
|
|
|||
There was a problem hiding this comment.
What is the purpose of the changes to this function? It seems you replace it in your subclass anyway.
| converged = True | ||
| break | ||
| log.info ("LASSCF rdm-jk {}".format (('not converged','converged')[int(converged)])) | ||
| return e_cas, casdm1frs, casdm2fr, converged | ||
|
|
||
| def kernel (las, mo_coeff=None, casdm1frs=None, casdm2fr=None, conv_tol_grad=1e-4, verbose=lib.logger.NOTE): |
There was a problem hiding this comment.
Editing this function changes the behavior of the existing lasscf_rdm implementation as well as your new implementation. Can you instead write a separate kernel function for your subclass, if you must change something in kernel?
There was a problem hiding this comment.
Or, alternatively, you could instead write "hook" functions that do nothing for the lasscf_rdm.LASSCFNoSymm class but which do the I/O for the lasscf_rdm.extremeAsynLASSCF class.
| file_path = os.path.join(directory_path, 'guessOrb.h5') | ||
| with h5py.File(file_path, 'r') as f: | ||
| mo_coeff = f['guessOrb'][:] | ||
| print("read the previous orbital") |
There was a problem hiding this comment.
We should not use "print". Use the logger commands instead.
| for it in range (las.max_cycle_macro): | ||
| e_cas, casdm1frs, casdm2fr, rdmjk_conv = rdm_cycle (las, mo_coeff, casdm1frs, | ||
| for it in range (las.max_cycle_macro): ###las.rdm_cycle | ||
| e_cas, casdm1frs, casdm2fr, rdmjk_conv = las.rdm_cycle (mo_coeff, casdm1frs, |
There was a problem hiding this comment.
This is kind of an example of a "hook" that my comment referred to above. las.rdm_cycle does one thing for lasscf_rdm.LASSCFNoSymm, and another thing for lasscf_rdm.extremeAsymLASSCF.
| @@ -503,6 +525,116 @@ def kernel(self, mo_coeff=None, casdm1frs=None, casdm2fr=None, conv_tol_grad=Non | |||
|
|
|||
| return self.e_tot, self.e_cas, self.casdm1frs, self.casdm2fr, self.mo_coeff, self.mo_energy, h2eff_sub, veff | |||
|
|
|||
| class extremeAsynLASSCF (LASSCFNoSymm): | |||
There was a problem hiding this comment.
At a certain point, if this gets too big, it would make sense to put it in a new file.
my_pyscf/mcscf/lasci_sync.py
Outdated
| @@ -4,7 +4,7 @@ | |||
| from scipy.sparse import linalg as sparse_linalg | |||
| from scipy import linalg | |||
| import numpy as np | |||
|
|
|||
| import h5py | |||
|
IDK if we're moving forward with this at this point but it might be a good exercise to respond to my comments and see if we can reach a good point with this PR, if we have time. |
|
Going to keep this branch separated (on my fork) while Mario and I continue working on it. |
