INVALID LANGUAGE PAIR SPECIFIED. EXAMPLE: LANGPAIR=EN|IT USING 2 LETTER ISO OR RFC3066 LIKE ZH-CN. ALMOST ALL LANGUAGES SUPPORTED BUT SOME MAY HAVE NO CONTENT > > | format | this model’s size | > |—|—:| > | f16 | 47.94 GB | > | Q80 | ~25.41 GB | > | Q6K | 19.68 GB | > | Q4KM | ~13.90 GB | > | PollardMix (this repo’s IQ1KT) | 5.82 GB** | Pollard builds of LiquidAI/LFM2.5-24B-A2B made with Pollard Weights — a ladder of measured-allocation quants (bits placed by per-layer sensitivity, not a uniform crush). Standard GGUF — runs in stock llama.cpp / ikllama.cpp, Ollama, LM Studio, except where noted. IQ1KT needs ikllama.cpp: its allocation puts ikllama-only atoms on the tensors it protects. The rest run anywhere. Every rung is the same weights, sized to a different RAM budget by the measured allocation. Pick the largest one that fits your machine with room for context: — ~21.7 GB RAM / VRAM -> Q6K (19.68 GB). — ~16.9 GB RAM / VRAM -> IQ4XS (14.85 GB). — ~13.5 GB RAM / VRAM -> IQ3S (11.53 GB). — ~10.9 GB RAM / VRAM -> IQ2XXS (8.89 GB). — ~7.8 GB RAM / VRAM -> IQ1KT (5.82 GB). (ikllama.cpp) Top-1 agree = share of tokens where the rung’s most-likely token is the same one the f16…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: PollardWeights
Теги: gguf, pollard-weights, pollard, llama.cpp, ik_llama.cpp, trellis, imatrix, mlx
Лайков: 3 | Загрузок: 2,393
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
