This model was asked to be published under my account, not the creators. The compute came from https://huggingface.co/posts/ProCreations/855858308074329

Boopit 1 Chat

Chat fine-tune of ProCreations/boopit-1: the same 1.58-bit ternary 28M / 4096-context model, SFT'd on mookiezi/Discord-Dialogues.

Packed weights stay under 7MB.

Prompt: <bos> your message <sep> and sample until <eos>.

Training

  • 200,000,215 assistant tokens, 383,910 steps
  • NVIDIA RTX PRO 6000 Blackwell
Downloads last month
112
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ProCreations/boopit-1-chat

Finetuned
(1)
this model

Datasets used to train ProCreations/boopit-1-chat