BorealisAI/flora-opt

This is the official repository for the paper "Flora: Low-Rank Adapters Are Secretly Gradient Compressors" in ICML 2024.

32
/ 100
Emerging

Training large AI models, especially large language models (LLMs), often requires immense GPU memory, which can be a significant bottleneck. Flora dramatically reduces the GPU memory needed for pre-training and fine-tuning these models without sacrificing performance or noticeably slowing down the process. AI/ML researchers and engineers working on deep learning model development can use this tool to develop and experiment with large models on more constrained hardware.

106 stars. No commits in the last 6 months.

Use this if you need to pre-train or fine-tune large deep learning models, such as LLMs, and are running into GPU memory limitations.

Not ideal if you are working with smaller models that do not push the limits of your GPU memory, as the benefits might not outweigh the integration effort.

deep-learning-training large-language-models gpu-optimization model-fine-tuning machine-learning-engineering
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 7 / 25

How are scores calculated?

Stars

106

Forks

5

Language

Python

License

LGPL-3.0

Last pushed

Jul 01, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/BorealisAI/flora-opt"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.