Struct DenseFp8Inputs 

pub struct DenseFp8Inputs {
    pub ffn_fp8: bool,
    pub attn_fp8: bool,
    pub keep_nvfp4: bool,
    pub dispatch: GemmDispatch,
    pub w8a8_kernels: bool,
    pub attn_w4a4: bool,
    pub attn_prefill_q_t: bool,
}

2026-09-25: The inputs of DenseFp8Plan::resolve. A plain struct, so the decision-table tests set every clause without the process environment, which keep_nvfp4_fallback caches in a OnceLock.

Fields

ffn_fp8: bool

2026-09-25: The native FP8 dense-FFN overlay is installed on this layer (qwen35_dense::ffn_fp8_arm_selected: METRALE_DENSE_FP8=1, tp size 1, Fp8Dequanted, a native FP8 gate_proj).

attn_fp8: bool

2026-09-25: The native FP8 attention overlay is installed on this layer.

keep_nvfp4: bool

2026-09-25: METRALE_DENSE_FP8_KEEP_NVFP4 (keep_nvfp4_fallback).

dispatch: GemmDispatch

2026-09-25: GemmDispatch::from_env(), the value the model stores as its dispatch.

w8a8_kernels: bool

2026-09-25: Both W8A8 prefill kernels are loaded (per_token_group_quant_fp8 and fp8_gemm_t_blockscaled). The W8A8 arms of prefill/paged_qkv.rs and prefill/paged_oproj.rs need both; without them those projections fall through to the transposed W8A16 kernels, which read the FP8 twins.

attn_w4a4: bool

2026-09-25: METRALE_ATTN_W4A4 is set. The W4A4 arm of prefill/paged_oproj.rs checks no weight type and reads self.attn.o_proj, the NVFP4 o_proj, so this lever keeps the NVFP4 attention weights under an FP8 overlay. The Q/K/V W4A4 arm in prefill/paged_qkv.rs requires an NVFP4 q_weight, so it never runs on an FP8 layer.

attn_prefill_q_t: bool

2026-09-25: METRALE_ATTN_PREFILL_Q_T=1. prefill/cache_skip_qkv.rs reads it on every call and, when set, lets Q use q_fp8w_t.

Trait Implementations

impl Clone for DenseFp8Inputs

fn clone(&self) -> DenseFp8Inputs

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DenseFp8Inputs

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for DenseFp8Inputs

fn eq(&self, other: &DenseFp8Inputs) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for DenseFp8Inputs

impl Eq for DenseFp8Inputs

impl StructuralPartialEq for DenseFp8Inputs