Add model_to_float32 / model_to_float64 model transforms - #8402

Draft

velochy wants to merge 4 commits into

pymc-devs:mainfrom

velochy:model-to-float32

Commits on Aug 17, 2026

Add model_to_float32 and model_to_float64 model transforms

velochy
and
claude
committed
99dacfa
Handle dtype aliases and foreign-dtype transform constants in model_to_float32

velochy
and
claude
committed
30de0af
Convert wrapped transform graphs instead of only casting their outputs

velochy
and
claude
committed
3fd4e6d
Calibrate model_to_float32 speed claims to measured CPU results

velochy
and
claude
committed
2d88df4