Layer-adaptive pruning produces models where some layers retain all experts (important layers) while others are aggressively pruned. moe-stream reads the per-layer expert count from the GGUF metadata ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results