Inkling-2-Mini-nvfp4 / conversion.json
prince-canuma's picture
Upload validated MLX Inkling 2 Mini with native mixed precision and bundled draft
26f71ee verified
Raw History Blame Contribute Delete
600 Bytes
{
"source_model": "thinkingmachines/Inkling-2-Mini",
"source_revision": "6e7c8248f75265bdd4fbab5273304243b5d8c2cb",
"mlx_vlm_base_revision": "392052b4439cf027ab42674d9977f62bb3dfa056",
"mlx_vlm_version": "0.8.0+inkling2mini",
"mlx_version": "0.32.4",
"transformers_version": "5.23.0",
"conversion": "MLX-native tensor renaming, gate/up row separation, and packed layout conversion; source expert precision retained.",
"target_weight_bytes": 19526043704,
"draft_weight_bytes": 635801416,
"runtime_patch_sha256": "e504b0e41d86939b1b1753abfb752151c64ceb9911df09eef83f8a1fabb42546"
}