Yahoo!天気 |地震天気JP 地震Yahoo!路線 |Excite辞書黄砂
Menu編集Wiki管理メモ
https://twitter.com/BoufrawFrodo2/status/1763435835935047789 オリジナルのBitNetを1.58bの論文に従って3値にするように修正しました
1.58 bit code is out (in an appendix) https://github.com/microsoft/unilm/blob/master/bitnet/The-Era-of-1-bit-LLMs__Training_Tips_Code_FAQ.pdf
It seems that results of that Microsoft paper about ternary LLMs can be replicated after all – for 3B@100B at least. https://huggingface.co/1bitLLM/bitnet_b1_58-3B