3 ms·
QuarterBit – Train 70B LLMs on a single GPU
- quarterbit 7mo agoI built QuarterBit because AI training costs are insane. A 70B model needs 840GB of memory — that's 11 A100 GPUs at $30+/hour. QuarterBit AXIOM compresses training memory 15x. Same model. Same quality. Fraction of the hardware. RESULTS: Llama 70B: 840GB → 53GB (11 GPUs → 1 GPU) = 90% savings Llama 13B: 156GB → 9GB (FREE on Kaggle T4) = 100% savings 91% energy reduction vs standard training. 100% trainable weights (not LoRA/adapters). 3 lines of code. HOW IT WORKS: from quarterbit import axiom model = axiom(model) model.cuda() TRY IT: pip install quarterbit Demo (FREE): https://www.kaggle.com/code/kyleclouthier/quarterbit-axiom-13b-demo-democratizing-ai https://www.kaggle.com/code/kyleclouthier/quarterbit-axiom-1... Benchmarks: https://quarterbit.dev https://quarterbit.dev AXIOM uses a novel weight representation combining lossless compression with a built-in optimizer. Weights stored at 0.62 bytes/param vs 4 bytes FP32. Gradient updates happen directly in compressed space. Not quantization-aware training or LoRA — every parameter fully trainable, convergence matches AdamW. Solo founder from Canada. Self-taught CUDA/ML. Applying to YC S26. Happy to answer questions.
- smallerize 7mo agoPut two spaces at the beginning of a line for monospace. Like this
- quarterbit 7mo agoThanks for the tip! It won't let me edit it. I think I will hide and repost it does look sloppy.
- quarterbit 7mo agoPlease download our Collab Blackwell training results on a 70b Llama model for review https://quarterbit.dev/data/QuarterBit_AXIOM_70B.ipynb https://quarterbit.dev/data/QuarterBit_AXIOM_70B.ipynb. This should be impossible and yet here it is.