Alexander Korolev e0eeffe3b1 lab-dflash2: use dequantized bf16 drafter variant (arch DFlashDraftModel for lab image; MXFP8 scales unsupported by its loader) 1 ay önce
..
recipe.toml e0eeffe3b1 lab-dflash2: use dequantized bf16 drafter variant (arch DFlashDraftModel for lab image; MXFP8 scales unsupported by its loader) 1 ay önce