Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@phases/00-setup-and-tooling/03-gpu-setup-and-cloud/code/mps_check.py`:
- Line 3: Remove the psutil import and any psutil-based RAM probes in this file
and replace them with a stdlib solution using os.sysconf; specifically, delete
the "import psutil" and modify the RAM-check logic (wherever
psutil.virtual_memory() or psutil.total is used) to call
os.sysconf('SC_PAGE_SIZE') * os.sysconf('SC_PHYS_PAGES') (or os.sysconf_names
variant) to compute total RAM, keeping the same variable names and checks so
functions like the module-level RAM check or any function that references the
previous psutil-derived value continue to work without changing external
behavior.
---
Nitpick comments:
In `@phases/00-setup-and-tooling/03-gpu-setup-and-cloud/code/mps_check.py`:
- Around line 9-11: Cache the result of torch.backends.mps.is_available() in a
local variable (e.g., mps_available) and reuse it for both the print and the
subsequent if-check instead of calling torch.backends.mps.is_available() twice;
update the print call and the if statement to reference mps_available to
simplify control flow and avoid repeated backend queries (apply same change in
the similar occurrence around lines 30-30).