Hi, Sungjin-

That amount of obliquity shouldn't matter at all-- I would not bother with deobliquing it.

Those voxels are standard+large, so I don't think those dsets are so huge.

The script that you attached (several hundred lines of commented beauty) was generated by a single AFNI command, "afni_proc.py". If you notice, at the top of that script, it says this:

echo "auto-generated by afni_proc.py, Tue Oct 15 15:34:29 2019"

So, I wanted the command used to *generate* that script; I am noticing now that the script itself contains the afni_proc.py command used to make it, stored in a comment at the bottom (Rick Reynolds, who wrote the afni_proc.py functionality, thought of *everything*); the command was (I have removed the comment "#" symbols):

afni_proc.py -subj_id 265001 -script proc.265001 -scr_overwrite -blocks \
 tshift despike align tlrc volreg blur mask scale regress -copy_anat \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/anatSS.SSW.nii \
 -anat_has_skull no -dsets \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.rest1.s1+orig.HEAD \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.pv.s1+orig.HEAD \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.14p.s1+orig.HEAD \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.6p.s1+orig.HEAD \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.flk.s1+orig.HEAD \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/deob.rest2.s1+orig.HEAD \
 -tcat_remove_first_trs 0 -align_opts_aea -cost lpc+ZZ -volreg_align_to \
 MIN_OUTLIER -volreg_align_e2a -volreg_tlrc_warp -tlrc_base \
 MNI152_2009_template_SSW.nii.gz -tlrc_NL_warp -tlrc_NL_warped_dsets \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/anatQQ.SSW.nii \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/anatQQ.SSW.aff12.1D \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Session1/265001_S1/subject_raw/anatQQ.SSW_WARP.nii \
 -blur_size 6.0 -regress_stim_times \
 /home/sungjin/fMRI/CGE/CGE_Raw_Data/Timing_CGE.txt -regress_stim_labels \
 Timing_CGE.txt -regress_basis 'BLOCK(6,1)' -regress_censor_motion 0.5 \
 -regress_apply_mot_types demean deriv -regress_motion_per_run \
 -regress_opts_3dD -jobs 4 -GOFORIT 5 -regress_reml_exec \
 -regress_compute_fitts -regress_make_ideal_sum sum_ideal.1D \
 -regress_est_blur_epits -regress_est_blur_errts

I don't see anything that particularly jumps out at me; I guess the REML part is somewhat memory intensive. I will ping Rick to see if he has any ideas on the memory aspect.

-pt