memory multiplier of input

[jdkent]

I'm curious what the peak memory consumption for an input file is. If a bold file is 1gb, will the peak memory being 2gb/5gb/10g? Are there large variables being held in memory longer than they need to be?

[htwangtw]

This is a great question and I would consider this an upstream issue in nilearn.
See this relevant issue on memory profiling of masker: nilearn/nilearn#3399
The memory consumption will depend on

1. size of input file
2. atlas type (segregated or probability)
3. did the masker object performs resampling

I consider this an upstream issue, as the current software depends on the nilearn masker object