GaLore: Advancing Large Model Training on Consumer-grade Hardware
β’
25
Titus von Koeller
Titus-von-Koeller
AI & ML interests
NN Quantization, Generative AI, LLMs, alignment, algorithms for social justice, ethical humanism, mitigating gender bias, audio compression, AGI
Articles
Organizations
Posts
3
Post
1906
π₯ Level up your model training w/ GaLore + Transformers for SOTA results on consumer-grade hardware!
β¬οΈ 82.5% less optimizer state memory footprint without performance degradation by expressing the gradient weight matrix as low rank.
π©πΏβπ» Install via
The integration of GaLore into the training of large language models (LLMs) marks a significant advancement in the field of deep learning, particularly in terms of memory efficiency and the democratization of AI research. By allowing for the training of billion-parameter models on consumer-grade hardware, reducing memory footprint in optimizer states, and leveraging advanced projection matrix techniques, GaLore opens new horizons for researchers and practitioners with limited access to high-end computational resources.
π¬ Find out more about GaLore and investigate lots of juicy technical details: https://huggingface.co/blog/galore
π€ Huge thanks to everyone involved β€οΈ:
β’ authors: @jiaweizhao @Kyriection @beidic Zhangyang Wang @animakumar @tydsh
β’ community contributors: @hiyouga @mdouglas and others!
β’ @ybelkada for taking such swift action in composing and coordinating necessary PRs to get this live at β‘ speed!
ποΈπ Super rewarding to see how @timdettmers work with optimizers is being built upon to achieve even greater heights!
π§ Actually, there are ongoing works to integrate GaLore into bitsandbytes and optimize memory efficiency even further πͺ. We'll keep you posted!
β¬οΈ 82.5% less optimizer state memory footprint without performance degradation by expressing the gradient weight matrix as low rank.
π©πΏβπ» Install via
pip install transformers>=4.39.0 galore-torch. #ProudlyGpuPoorThe integration of GaLore into the training of large language models (LLMs) marks a significant advancement in the field of deep learning, particularly in terms of memory efficiency and the democratization of AI research. By allowing for the training of billion-parameter models on consumer-grade hardware, reducing memory footprint in optimizer states, and leveraging advanced projection matrix techniques, GaLore opens new horizons for researchers and practitioners with limited access to high-end computational resources.
π¬ Find out more about GaLore and investigate lots of juicy technical details: https://huggingface.co/blog/galore
π€ Huge thanks to everyone involved β€οΈ:
β’ authors: @jiaweizhao @Kyriection @beidic Zhangyang Wang @animakumar @tydsh
β’ community contributors: @hiyouga @mdouglas and others!
β’ @ybelkada for taking such swift action in composing and coordinating necessary PRs to get this live at β‘ speed!
ποΈπ Super rewarding to see how @timdettmers work with optimizers is being built upon to achieve even greater heights!
π§ Actually, there are ongoing works to integrate GaLore into bitsandbytes and optimize memory efficiency even further πͺ. We'll keep you posted!
Post
We just released bitsandbytes==0.43.0 π¦ , with these significant new additions:
β£ π« FSDP+QLoRA support (alpha release)
β¦ now anyone with 2 powerful gaming GPUs can fine-tune 70B param models at home!
β¦ in collab with Jeremy Howard + team @ answer.ai
β¦ answer.ai blogpost: https://www.answer.ai/posts/2024-03-06-fsdp-qlora.html
β¦ example repo: https://github.com/AnswerDotAI/fsdp_qlora/
β£ πβ Official Windows support
β¦ now via simple
β£ π Huge docs update:
β¦ https://huggingface.co/docs/bitsandbytes/main
β¦ Be sure to check out the optimizers and the API docs
β¦ ... even more upcoming ...
Under the hood there we have many other improvements, due to extensive maintenance activity, community contributions by super active + knowledgable volunteers β¨ π and the official sponsorship by Hugging Face that makes all this possible π€ β€οΈ π
We would greatly appreciate any further community contributions, be it to help with refactorings, exterminating flaky tests, writing doc-strings, tutorials, new features. Don't be shy, just contact us and we see where this leads us:
https://github.com/TimDettmers/bitsandbytes/discussions
Have a great weekend everyone!
β£ π« FSDP+QLoRA support (alpha release)
β¦ now anyone with 2 powerful gaming GPUs can fine-tune 70B param models at home!
β¦ in collab with Jeremy Howard + team @ answer.ai
β¦ answer.ai blogpost: https://www.answer.ai/posts/2024-03-06-fsdp-qlora.html
β¦ example repo: https://github.com/AnswerDotAI/fsdp_qlora/
β£ πβ Official Windows support
β¦ now via simple
pip install bitsandbytes>=0.43.0β£ π Huge docs update:
β¦ https://huggingface.co/docs/bitsandbytes/main
β¦ Be sure to check out the optimizers and the API docs
β¦ ... even more upcoming ...
Under the hood there we have many other improvements, due to extensive maintenance activity, community contributions by super active + knowledgable volunteers β¨ π and the official sponsorship by Hugging Face that makes all this possible π€ β€οΈ π
We would greatly appreciate any further community contributions, be it to help with refactorings, exterminating flaky tests, writing doc-strings, tutorials, new features. Don't be shy, just contact us and we see where this leads us:
https://github.com/TimDettmers/bitsandbytes/discussions
Have a great weekend everyone!
models
None public yet
datasets
None public yet