BitNet: Scaling 1-bit Transformers for Large Language Models
1 min read
The increasing size of large language models has posed challenges for deployment and raised concerns about environmental impact due to high energy consumption. In this work, we introduce BitNet, a scalable and stable 1-bit Transformer architecture designed for large language models. Specifically, we introduce BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Experimental results on language modeling show that BitNet achieves competitive performance while substantially reducing memory footprint and energy consumption, compared to state-of-the-art 8-bit quantization methods and FP16 Transformer baselines. Furthermore, BitNet exhibits a scaling law akin to full-precision Transformers, suggesting its potential for effective scaling to even larger language models while maintaining efficiency and performance benefits.
Shaun’s 2c FWIW: Picture this, a very efficeient 1-bit LLM that Runs Directly on CPUs (not GPUs or dedicated NPUs). Microsoft recently open-sourced bitnet.cpp, and even large 100-billion parameter models can be executed on local devices without the need for special hardware.
