Repository navigation

[pyTorch] CPU performance optimizations - #2439

Merged

ptrendx merged 26 commits into

NVIDIA:mainfrom

ptrendx:pr_python_cpu_optimization

Jan 21, 2026

Merged

[pyTorch] CPU performance optimizations#2439
ptrendx merged 26 commits into
NVIDIA:mainfrom
ptrendx:pr_python_cpu_optimization

Commits on Jan 13, 2026
- PoC of the changes
  ptrendx
  committed
  c778208
- Early exit from the Free function for the empty tensor
  ptrendx
  committed
  1b1c574
- Use the proper function for nvtx range
  ptrendx
  committed
  ac14119
- Only do mark_not_offload when the cpu_offloading is enabled
  ptrendx
  committed
  616af19
- First pass on making the setattr issue not come back
  ptrendx
  committed
  c0d2ccc
- Actually add pytest.ini
  ptrendx
  committed
  68b6f74
- Changes to __init__
  ptrendx
  committed
  1c5434c
- A different way
  ptrendx
  committed
  778019d
- WAR the fact that it is not possible to set __setattr__ dynamically
  ptrendx
  committed
  0fc2a62
- Simpler solution and fixes
  ptrendx
  committed
  2fb6ee3
Commits on Jan 14, 2026
- Fix for the inference mode DPA
  ptrendx
  committed
  4940724
- Start of debugging debug tools
  ptrendx
  committed
  8704a59
- More fixes in debug
  ptrendx
  committed
  aa783a7
- [pre-commit.ci] auto fixes from pre-commit.com hooks
  pre-commit-ci[bot]
  committed
  fd158f6
- Speculative moving the validate_name to the constructor
  ptrendx
  committed
  2601e5e
- [pre-commit.ci] auto fixes from pre-commit.com hooks
  pre-commit-ci[bot]
  committed
  09a3a7c
- Fix
  ptrendx
  committed
  026b4b1
- Making the debug tools names saner
  ptrendx
  committed
  e83e816
- Change the setattr usage in the tensor parallel group setting
  ptrendx
  committed
  8843a46
- Adding try/finally - it does not seem to impact the time in observable
  ptrendx
  committed
  084847f
Commits on Jan 15, 2026
- Fixing lint issues and the thunder test
  ptrendx
  committed
  d004747
- Fix 1 of the debug tests
  ptrendx
  committed
  523944a
- Removed the warning and enforcement in the CI
  ptrendx
  committed
  9873235
Commits on Jan 16, 2026
- try-finally in the context manager
  ptrendx
  committed
  7b55639
Commits on Jan 20, 2026
- Fixing the debug tests
  ptrendx
  committed
  028d03f
- Merge branch 'main' into pr_python_cpu_optimization
  ptrendx
  authored
  0d949ef