NVFP4_GROUPED_DECODE_TC_MAX_ROWS

metrale_model_layers::layers::moe

Constant NVFP4_GROUPED_DECODE_TC_MAX_ROWS 

Source 
pub const NVFP4_GROUPED_DECODE_TC_MAX_ROWS: usize = super::forward_fp8_grouped_decode::FP8_GROUPED_DECODE_TC_MAX_ROWS; // 256usize
Expand description

2026-10-02: Widest row count on the tensor-core expert kernels, the grouped FP8 tensor-core decode’s: one verify of 128 sequences at one draft.