MPT-30B: Raising the bar for open-source foundation models
www.mosaicml.com
Introducing MPT-30B, a new, more powerful member of our Foundation Series of open-source models, trained with an 8k context length on H100s.
  • BehohippyEnglish
    3•1 year ago

    Still had some reasoning issues, but looking forward to the fine tunes!

    • @[email protected]OPMEnglish
      2•1 year ago

      Does prompting in a way to improve it’s reasoning work? I’m talking about CoT or ToT or other prompting strategies.

      • BehohippyEnglish
        1•1 year ago

        CoT maybe, but ToT for sure. I tried the same thing on vicuna-1.1-13b and it got it after 2 prompts with ToT