keep_nvfp4_fallback

spark_model::weight_loader::fp8_residency

Function keep_nvfp4_fallback 

Source 
pub fn keep_nvfp4_fallback() -> bool
Expand description

The per-layer kill switch that restores the pre-#915 behaviour: build every NVFP4 fallback copy even where the dispatch cannot reach it.

PRESENCE (any value, including empty), matching ATLAS_FFN_W8A16_ONLY — this is an escape hatch an operator reaches for while a serve is misbehaving, and ATLAS_DENSE_FP8_KEEP_NVFP4=0 meaning “on” is a trap.