[torchlib] Simplify linalg_vector_norm to remove the redundant Abs

[justinchuby]

This happens in some of the LORA models. When we use ReduceL1/ReduceL2 or when ord is an even number, we don't need to take Abs of the input

Signed-off-by: Justin Chu justinchuby@users.noreply.github.com

[Copilot]

Pull Request Overview

This PR optimizes the aten_linalg_vector_norm function by selectively applying the absolute value operation only when necessary, based on the order parameter. The change removes redundant Abs operations for even-order norms and built-in reduction operations.

• Moves the Abs operation from being applied unconditionally to being conditionally applied based on the norm order
• Adds logic to only apply Abs for negative orders, odd orders, or infinite orders
• Maintains existing behavior for ReduceL1, ReduceL2, and even-order norms without unnecessary absolute value computation

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 70.09%. Comparing base (f54cf47) to head (5a8e7ee).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2570   +/-   ##
=======================================
  Coverage   70.09%   70.09%           
=======================================
  Files         222      222           
  Lines       26182    26184    +2     
  Branches     2580     2581    +1     
=======================================
+ Hits        18351    18353    +2     
  Misses       6931     6931           
  Partials      900      900           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.