Tether announces QVAC, a cross-platform BitNet LoRA framework: enabling training of billion-parameter AI models on consumer-grade devices

Odaily Planet Daily reports that according to an official announcement, Tether has launched a cross-platform BitNet LoRA fine-tuning framework within QVAC Fabric, optimized for training and inference of Microsoft BitNet (1-bit LLM). This framework significantly reduces computational power and memory requirements, enabling billion-parameter models to be trained and fine-tuned on laptops, consumer GPUs, and smartphones.

This is the first time that the BitNet model has been fine-tuned on mobile GPU devices (including Adreno, Mali, and Apple Bionic). Tests show that a 125M parameter model can be fine-tuned in about 10 minutes, a 1B model in roughly an hour, and even scaled up to a 13B parameter model on smartphones.

Additionally, the framework supports heterogeneous hardware such as Intel, AMD, and Apple Silicon, and for the first time achieves 1-bit LLM LoRA fine-tuning on non-NVIDIA devices. In terms of performance, BitNet models infer 2 to 11 times faster on mobile GPUs compared to CPUs, while reducing VRAM usage by up to approximately 77.8% compared to traditional 16-bit models.

Tether states that this technology is expected to break dependence on high-end computing power and cloud infrastructure, promote decentralized and localized AI training, and lay the foundation for new applications like federated learning.