423
DeepSeek ditches Nvidia for Huawei chips in V4 launch (cybernews.com)
submitted 6 months ago by to c/technology@lemmy.world
89 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 2 points 6 months ago* (2 children)

Not to speak of the new attention scheme and the (IIRC) MLP changes.

I’m very much looking forward to ik_llama.cpp implementing it. I don’t think I can quite fit Flash on my rig (hence no Ktransformers for me) but a little quantization of the sparse layers, and it’d be perfect.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 6 months ago (1 child)

    Makes sense, even Flash is fairly sizable! KTransformers also has a "llamafile" backend which uses GGUFs, but ik_llama will almost certainly perform better if you're not on a NUMA setup. In my case, I'm using a dual socket motherboard, so KTransformers performs quite a bit better (I think ik_llama hasn't implemented extensive NUMA optimizations quite yet, but sounds like it's coming), though I normally use KTransformers for native FP8 weights.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 6 months ago

    It is! 143GB last I checked. I'm on 128GB RAM + 3090, 1 NUMA node, so I think it's juuust barely too tight. But it should be perfect with a few of the "sparsest" MoEs quantized.

    If KTransformers supports something like that, I may have to finally check it out, since v4 won't need many esoteric features.

  • permalink
  • fedilink
  • source
  • parent