Movie generation ligand protein clashes

[Jnelen]

I'm seeing some odd results when generating a movie from my results. First of all, the ligand seems to be clashing with the protein backbone in the first few steps.
Additionally, I was curious regarding the last 2 steps since it doesn't seem to do any relaxing anymore? A lot of code here is commented away, and before there seems to have been a openmm_relax function.

if step < inference_steps-1:
            relaxed_ligandFile = ligandFile
            try:
                retry = 0
                ret = openmm_relax_protein_only((pdbFile, fixed_pdbFile, relaxed_proteinFile, gap_mask, stiffness, use_gpu))
                while ret['efinal'] > 0 and retry < 5:
                    # if ret['einit'] > 0 and ret['efinal'] / ret['einit'] < 0.1:
                    #     break
                    ret = openmm_relax_protein_only((relaxed_proteinFile, fixed_pdbFile, relaxed_proteinFile, gap_mask, stiffness, use_gpu))
                    print(ret)

                    retry += 1
                # print(ret['einit'],ret['efinal'])
            except Exception as e:
                print(e, "relax fail, use original instead")
                _ = os.system(f"cp {pdbFile} {relaxed_proteinFile}")
else:
            relaxed_ligandFile = os.path.join(write_dir, [fn for fn in file_names if f'rank{rank}_ligand_lddt' in fn and 'relaxed' in fn][0])
            relaxed_proteinFile = os.path.join(write_dir, [fn for fn in file_names if f'rank{rank}_receptor_lddt' in fn and 'relaxed' in fn][0])
            # relaxed_ligandFile = os.path.join(write_dir, f'rank{rank}_ligand_step{step+1}_relaxed.sdf')

You can find an example output here: Test_reverseprocess.zip. Is this expected behavior, or did I do something wrong? Some clarification would be greatly appreciated!