Files
zeekayandHanzo Dev b639a2b5f9 engine: a gathered-prefix prompt is causal unless its FlashParams say otherwise
The packed varlen arm defaulted a missing FlashParams to "the mask says", so a quantized model's later
chunk (no FlashParams, no mask) ran the fused kernel bidirectional. None means a causal decoder, as
run_attention documents. The mask helper now keys on whether a fused varlen kernel will run, not on
the device, so CUDA with flash off or an F32 model keeps its mask too.

Co-authored-by: Hanzo Dev <[email protected]>
2026-09-18 18:39:09 -07:00
..

hanzo-engine

Core crate of hanzo including the models and associated executors.

Documentation: https://docs.rs/hanzo-engine/