github-featured / ollama--ollama

发布

0 发布 616 Git 标签

RSS 订阅源
  • v0.31.1 710292ff4f
    比较

    mlx: tighten up gemma4 moe loading code (#16964)

    frostbyte_neo 发布于 2026-06-30 04:15:08 +00:00 | 101 次提交 在此版本后已推送到 main

    This change allows .experts.gate_proj / .up_proj / .down_proj tensor names to each
    be used for both quantized (i.e. nvfp4 and mxfp8) and non-quantized (bf16) models.
    Previous to this only non-quantized models used that tensor naming scheme.

    下载附件