numerics_debugging

Installation
SKILL.md

Numerics Debugging (DebugMode-based)

Per-op activation capture + comparison toolkit. Captures activations on a designated step via torch.utils._debug_mode.DebugMode, then diffs two captures into an HTML report to surface numerics divergence between runs that should agree (bitwise, or within float32 reduction-order noise).

Two pieces live in scripts/:

  • activation_tracer.py — runtime capture, driven from Profiler by ActivationCaptureProfiler. Output: {dump_folder}/numerics/rank_{N}_activations.log.
  • compare_numerics.py — diffs two logs, produces an HTML report. Standard library only, no torch import.
Installs
1
GitHub Stars
5.9K
First Seen
Oct 22, 2026