Bitnet again?
Summary
Microsoft has released a new embedding model called bitnet-embedding-0.6b on HuggingFace, but the model card is missing.
Similar Articles
BitNet Text Embeddings
This paper introduces BitEmbed, an extreme low-bit framework for LLM-based text embeddings that converts pretrained LLM backbones into BitNet-style encoders with ternary weights and quantized activations. It achieves comparable performance to full-precision models while significantly reducing encoding and storage costs.
NEW BITNET MODELS!
New BitCPM4-CANN models (1B, 3B, 8B) from OpenBMB released on Hugging Face; awaiting llamacpp support for testing.
@QuixiAI: @MicrosoftAI never published a BitNet trainer. I fixed that bug. https://github.com/QuixiAI/bitnet trained my own BitNe…
A user released a BitNet trainer that Microsoft never published, along with custom kernels for training and inference, while also highlighting Microsoft's bitnet.cpp inference framework for fast 1-bit LLM inference on CPUs and GPUs.
Bitnet.cpp: Efficient Edge Inference for Ternary LLMs
Bitnet.cpp presents a mixed-precision matrix multiplication library for efficient edge inference of ternary LLMs like BitNet b1.58, achieving up to 6.25x speedup over full-precision baselines. The system is open-sourced on GitHub.
Was BitNet a dead end? What happened to ternary LLMs?
The article questions why ternary language models like BitNet have not scaled beyond 2B parameters, given their initial promise, and discusses the apparent lack of progress from open-weight AI labs.