Thump604/DeepSeek-V4-Flash-MLX-Q3-mixed-gs128-affine

— Base model: deepseek-ai/DeepSeek-V4-Flash — Source revision: 6e763230a9d263eca2023f1d4a5ce1bfe126cf48 — Architecture: DeepseekV4ForCausalLM — Model type: deepseekv4` — Tooling branch: Thump604/mlx-lm, branch deepseek-v4-support-fixes — Minimum tooling commit for generation: 9c990f4 — Output path during conversion: /Volumes/Lexar/mlxmodels/DeepSeek-V4-Flash-MLX-Q3-mixed-gs128-affine — Quantization recipe: mixed36 — Quantization mode: affine — Group size: 128 — Effective bits — Conversion completed successfully. — Lazy MLX load completed successfully on a 128GB Mac Studio. — One-token generation smoke was attempted and stopped after memory pressure and swap activity exceeded the local safety boundary. Treat this artifact as converted and load-validated, not generation-qualified on 128GB Apple Silicon. DeepSeek V4…

Модальности:
Генерация текста

Задача: Генерация текста
Автор: Thump604
Теги: mlx, deepseek_v4, deepseek-v4, quantized, en, 4-bit
Лайков: 4  |  Загрузок: 344

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.