Repository navigation

Fix stale GGML fake-quant cache after version-bypassing optimizer steps (#2857) - #2874

Open

dcaayushd wants to merge 6 commits into

NVIDIA:mainfrom

dcaayushd:fix/ggml-stale-cache-2857

Open

Fix stale GGML fake-quant cache after version-bypassing optimizer steps (#2857)#2874
dcaayushd wants to merge 6 commits into
NVIDIA:mainfrom
dcaayushd:fix/ggml-stale-cache-2857

Commits on Oct 19, 2026
- Fix stale GGML fake-quant cache after version-bypassing optimizer steps (#2857)
  
  dcaayushd
  and
  Copilot
  committed
  ed33d6e
- Preserve GGML cache layout checks for export
  
  dcaayushd
  and
  Copilot
  committed
  24fd416
- Fix GGML cache validation during no-grad evaluation
  dcaayushd
  committed
  ae82e12
Commits on Oct 20, 2026
- Merge upstream/main into fix/ggml-stale-cache-2857
  dcaayushd
  committed
  b536fcb
- Merge branch 'main' into fix/ggml-stale-cache-2857
  dcaayushd
  authored
  c437e18
- Fix GGML cache validation for reshaped weights
  dcaayushd
  committed
  b3fbc93