Metraipolis › archive › dc_c5a0eb792dec

Every Expert Counts: ExactMoE for Memory-Efficient W4A16 Inference — arXiv:2608.15383

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.15383
License (as archived)
CC0-1.0
Archived
2026-10-18T19:40:18.713Z
SHA-256 of the archived text
28d804c710d6c592809fda47312f24b2563fde60b57257a7336b95e1e13ea740
Stable URL of this page
https://metraipolis.com/documento/dc_c5a0eb792dec

How to cite

Every Expert Counts: ExactMoE for Memory-Efficient W4A16 Inference — arXiv:2608.15383 — arXiv (en), https://arxiv.org/abs/2608.15383 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c5a0eb792dec (Metraipolis archive dc_c5a0eb792dec, sha256 28d804c710d6c592809fda47312f24b2563fde60b57257a7336b95e1e13ea740)
BibTeX
@misc{dcc5a0eb792dec,
  title = {Every Expert Counts: ExactMoE for Memory-Efficient W4A16 Inference — arXiv:2608.15383},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.15383},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c5a0eb792dec; sha256 28d804c710d6c592809fda47312f24b2563fde60b57257a7336b95e1e13ea740},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15383. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Sparse mixture-of-experts (MoE) language models reduce arithmetic by activating only a small subset of experts per token, yet deployment still requires storing and moving the full expert bank. We present ExactMoE, an inference design that applies symmetric group-128 four-bit weight quantization only to routed experts, stores those experts in kernel-native MARLIN form in pinned host memory, and executes all selected experts through a configurable GPU-resident slot cache and fused grouped MoE kernels. The router, attention, embeddings, normalization layers, and language-model head remain in BF16. "Exact" refers to complete expert availability and an unchanged top-k routing procedure: no expert is pruned, substituted, or forced to execute on the CPU. It does not imply numerical identity with the BF16 model. On OLMoE-1B-7B-0924-Instruct, evaluated on a single NVIDIA L4, a 16-slot configuration reduces peak reserved GPU memory from 14.168 to 1.836 GiB (87.04%) while retaining 81.85% of BF16 decode throughput. A fully resident 64-slot configuration reaches 31.923 tokens/s versus 21.662 tokens/s for BF16 while reserving 4.061 GiB. Across 12,450 zero-shot multiple-choice questions, ExactMoE obtains 70.3534% normalized accuracy versus 70.8996% for BF16, retaining 99.23% of the baseline accuracy. In a matched 16-token ablation, fused grouped execution is 1.97x as fast as a sequential W4 reference. These results identify a practical memory-transfer-throughput frontier for complete-expert MoE inference.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.15383 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-18T19:02:44.089Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).