The first MTPLX build of Qwopus3.6-27B-Coder — Jackrong’s agentic-coding fine-tune of Qwopus3.6-27B-v2 (repo-level coding, multi-turn tool orchestration, 67.0% SWE-bench Verified in no-thinking mode), with native multi-token-prediction speculative decoding on Apple Silicon. No external drafter, exact rejection sampling. 4-bit sibling: nom666/Qwopus3.6-27B-Coder-MTPLX-4bit-Speed. Forged with mtplx forge build from the original BF16 Jackrong/Qwopus3.6-27B-Coder: — Body: flat 8-bit MLX affine quantization, group size 64 — MTP head: preserved in BF16 (mtppolicy: keepbf16), packed as mtp.safetensors sidecar — Size: ~29 GB — Calibrated MTP contract included (mtplxruntime.json`) The Coder’s MTP head was trained with draft=2, which shows: depth-2 acceptance is exceptional and depth 2 beats depth 3. Verification suite: long-code-uncapped, 2048-token budget. Use —depth 2 when serving. Serves OpenAI-compatible (/v1/chat/completions) and Anthropic-compatible (/v1/messages) endpoints with warm-prefix KV reuse, SSD session cache, continuous batching, and vision support. Full 262144-token context; only 16 of 64 layers carry KV (hybrid Gated DeltaNet architecture). — This fine-tune targets…
Модальности:
Генерация текста
Области применения:
Генерация кода Диалог / чат
Задача: Генерация текста
Автор: nom666
Теги: mlx, qwen3_5, mtplx, mtp, speculative-decoding, apple-silicon, qwopus, code
Лайков: 3 | Загрузок: 53
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
