[Bug]: Invalid eigenvalues

[ioanaapapa]

Bug Report

When running a calculation with the current version of CodeEntropy on main, a lot of invalid eigenvalues are identified. This was not the case with an earlier version of CodeEntropy 2.1.1.

---

To Reproduce

YAML configuration

│ top_traj_file │ ['ramp2.prmtop', 'ramp2_1_500fr_imaged.nc']
│ force_file │ None │
│ file_format │ None │
│ kcal_force_units │ False │
│ selection_string │ all │
│ start │ 0 │
│ end │ -1 │
│ step │ 1 │
│ bin_width │ 30 │
│ temperature │ 298.0 │
│ verbose │ False │
│ grouping │ molecules │
│ combined_forcetorque │ False │
│ customised_axes │ False │
│ search_type │ RAD |

CLI command used

CodeEntropy --top_traj_file ramp2.prmtop ramp2_1_500fr_imaged.nc --combined_forcetorque False --customised_axes False --end -1

Actual output

program_output.txt

Expected output

Output was expected to be similar to that from a previous version, such as the one below.
program_output_prev.txt

Environment and Versions

• Operating System: Ubuntu 24.04.1
• CodeEntropy Version: 2.1.1
• Python Version: 3.12.0
• Package list: (https://github.com/user-attachments/files/26925992/packages.txt)

Additional Context

Issue appears with flag --customised_axes set either True or False.

[jimboid]

Thanks Ioana. @skfegan @harryswift01 we should check if this shows up in the systems used for testing and make sure we are testing for this. If it does not then we need to write a new set of tests to validate against this type of bug.

[skfegan]

Does this also happen with combined_forcetorque set to True?

[ioanaapapa]

It still happens with combined_forcetorque True

program_output.txt

[skfegan]

@jimboid @harryswift01 The regression tests [frame-window case] will catch this if it comes back, but this bug crept in when I was changing the trajectory slicing which affected the conformational and orientational results, so the regression baselines were updated with the wrong vibrational results for cases with more than one frame.