PR #2469 — docs: clarify canonical pruning documentation source (#1871)
Repo: NVIDIA/Model-Optimizer — 3.8k ⭐, unified library for SOTA
model optimization: quantization, distillation, pruning, NAS,
speculative decoding.
The problem I fixed:
Pruning docs were split between:
-
docs/source/guides/3_pruning.rst examples/pruning/README.md
No clear signal on which was authoritative — the README is the real
canonical source (covers Minitron, Puzzletron, FastNAS, support matrix,
distillation hyperparameters). I added a note to the RST guide making
that explicit.
Simple fix. But clarity in ML tooling docs matters — wrong docs =
wrong model compression decisions.
→ PR #2469
