LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Schultheis, Erik, Alistarh, Dan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918252178833408
author Schultheis, Erik
Alistarh, Dan
author_facet Schultheis, Erik
Alistarh, Dan
contents We present LLMQ, an end-to-end CUDA/C++ implementation for medium-sized language-model training, e.g. 3B to 32B parameters, on affordable, commodity GPUs. These devices are characterized by low memory availability and slow communication compared to datacentre-grade GPUs. Consequently, we showcase a range of optimizations that target these bottlenecks, including activation checkpointing, offloading, and copy-engine based collectives. LLMQ is able to train or fine-tune a 7B model on a single 16GB mid-range gaming card, or a 32B model on a workstation equipped with 4 RTX 4090s. This is achieved while executing a standard 8-bit training pipeline, without additional algorithmic approximations, and maintaining FLOP utilization of around 50%. The efficiency of LLMQ rivals that of production-scale systems on much more expensive cloud-grade GPUs.
format Preprint
id arxiv_https___arxiv_org_abs_2512_15306
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs
Schultheis, Erik
Alistarh, Dan
Distributed, Parallel, and Cluster Computing
Machine Learning
We present LLMQ, an end-to-end CUDA/C++ implementation for medium-sized language-model training, e.g. 3B to 32B parameters, on affordable, commodity GPUs. These devices are characterized by low memory availability and slow communication compared to datacentre-grade GPUs. Consequently, we showcase a range of optimizations that target these bottlenecks, including activation checkpointing, offloading, and copy-engine based collectives. LLMQ is able to train or fine-tune a 7B model on a single 16GB mid-range gaming card, or a 32B model on a workstation equipped with 4 RTX 4090s. This is achieved while executing a standard 8-bit training pipeline, without additional algorithmic approximations, and maintaining FLOP utilization of around 50%. The efficiency of LLMQ rivals that of production-scale systems on much more expensive cloud-grade GPUs.
title LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs
topic Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2512.15306