What is everyone’s experiences so far with DPO trained versions of their favorite models? Been messing around with different models and my two new favorite models are actually just the DPO versions of my previous favorite models (causalLM 14b and openhermes 2.5 7b). Links below for the models in question.

CausalLM 14B-DPO-alpha - GGUF: https://huggingface.co/tastypear/CausalLM-14B-DPO-alpha-GGUF

NeuralHermes 2.5 Mistral 7B - GGUF: https://huggingface.co/TheBloke/NeuralHermes-2.5-Mistral-7B-GGUF

The former runs at 30 t/s for me with koboldcpp-rocm on a 6900 XT, and the latter at 15 t/s, both at Q6K. I don’t have a favorite between these two models, they seem to be better at different things and trade blows in all the logic + creative writing tasks I’ve tested them in, despite causalLM being a larger model. I’m looking forward to seeing what nousresearch/teknium and CausalLM are bringing next.

  • 1monster90@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    Out of topic here but is it just me or open hermes somehow speaks in russian unprovoked frequently?

    • xadiant@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      High repetition penalty? One model I merged suddenly started speaking Spanish in one summarisation task lol

    • ttkciar@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      If you are using llama.cpp, you might want to give it a grammar which forces ASCII output.

    • VertexMachine@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      Interesting. I’m using oobabooga and that never happened to me. I actually don’t recall it ever outputting anything but English…