BitNet (Scaling 1-bit Transformers for Large Language Models)

BitNet | Microsoft

• https://www.microsoft.com/en-us/research/publication/bitnet-scaling-1-bit-transformers-for-large-language-models/
• https://arxiv.org/abs/2310.11453
  • https://arxiv.org/pdf/2310.11453.pdf

• https://github.com/microsoft/unilm/blob/master/bitnet/The-Era-of-1-bit-LLMs__Training_Tips_Code_FAQ.pdf

実装†

• https://github.com/microsoft/BitNet
• https://github.com/kyegomez/BitNet
• https://github.com/shi3z/BitNet

1.58 Bits†

• BitNet/1.58

.†

• BitNet
  • https://qiita.com/fuyu_quant/items/fa4fc3930ddfd2ec6f16

    ※BitNetそのものを検証した記事ではなく，BitNetで使われているレイヤーを使った簡単なNNモデルを作成し検証しています．