New FastCSP release: update to docs/example/configs/correctors - #2076
Conversation
| input_df["correction.e_fragments_corrector"] = sum_corr | ||
| input_df["correction.e_fragments_corrector_per_molecule"] = sum_corr / z_arr | ||
| input_df["energy_corrected"] = e_relaxed - sum_orig + sum_corr | ||
| input_df["energy_corrected_per_molecule"] = input_df["energy_corrected"] / z_arr |
There was a problem hiding this comment.
is the corrected energy ever used in filter? It looks like it will be used when removing problematic structures here (using "validity.conformer_corrections.applied") here:
fairchem/src/fairchem/applications/fastcsp/core/workflow/filter.py
Lines 271 to 274 in 784e86b
But is "energy_corrected_per_molecule" every used for ranking, filtering or free energy? Looks like only "energy_relaxed_per_molecule" is used.
There was a problem hiding this comment.
Good point. It was not but one can make it to ;) If smth is wrong with the energy correction validity calculations, we don't want to use it downstream anyway. We accounted exactly 0 such cases. As long as Z_check passes, correction is correctly computed.
No description provided.