Repository navigation

Widen float16/bfloat16 to float32 in cpu reduce - #4387

Merged

zcbenz merged 5 commits into

ml-explore:mainfrom

Ved235:variance-bug

Sep 1, 2026

Commits on Aug 30, 2026
- Fix floating point promotion in arithmetic reduction
  Ved235
  committed
  6aae9d6
- Add new test for bf16
  Ved235
  committed
  39fcd3d
- Add comment for widening to float32
  Ved235
  committed
  9840bce
- Remove widening to double
  Ved235
  committed
  8cfd696
- Remove stale widening_double reference
  Ved235
  committed
  0c9eb2b