The issue is that PEFT spends all that time doing unnecessarily doing kaiming init. So our code first patches the `init` module so that `kaiming_uniform_` does nothing at all. Here's a gist with the full code:
OPEN SOURCE
-
Speed up Hugging Face PEFT model initialization with this trick
By
–
If you use @huggingface PEFT, then you'll have noticed that calling `get_peft_model` is *really* slow… However, you can make it nearly instant by running this magic snippet first!
-
Small Team Creates Best 7B Model in Three Months
By
–
It's plenty enough. They created the best 7B model yet with a small team in 3 months. The size of an eng time is not generally positively correlated with its productivity.
-
Fast SDXL Inference with JAX on TPU v5e
By
–
Fast SDXL Inference with JAX on TPU v5e Amazing how fast anyone can generate 4 1024×1024 resolution SDXL images in under 5 seconds, for free: https://
huggingface.co/spaces/google/
sdxl
… -
XGBoost Community Celebrates Major Milestone with Global Contributors
By
–
Congrats to the XGBoost community and the contributors around the world who enabled this
-
Top talent prefers open source over closed development work
By
–
I think the absolute top talent are going to much prefer to work on open source over closed. There's a long history of the best people in the field publishing openly. Now that's an option again expect to see it become a recruiting draw card.
-
VC-backed companies increasingly adopt open source strategies
By
–
Many companies raising rounds / VC backed are also building open source to begin with which certainly wasn’t the case 5-10 years ago.
-
Open Source AI Tools Democratizing Engineering with Ludwig
By
–
The future is #opensource! Check out the new episode of Redefining #AI w/ host Lauren Hawker Zafer from Squirro as she sits down with @predibase co-founder @w4nderlus7 to discuss how tools like @ludwig_ai are democratizing AI for engineers. Check it out: https://
pbase.ai/3Q0Ggwh -
Fine-tuning Open Source LLMs: Simplify Deployment and Infrastructure
By
–
Customizing pretrained #opensource LLMs is top of mind for every #developer—but 70% say they struggle w/ #finetuning and managing #infra. Don't miss our upcoming webinar to learn how to deploy and fine-tune #LLMs in minutes on scalable infra in your VPC. https://
pbase.ai/3F248JI -
Stability AI Introduces Stable LM 3B for Smart Devices
By
–
Stability AI is proud to Introduce Stable LM 3B: Bringing Sustainable, High-Performance Language Models to Smart Devices. Download the weights on Hugging Face today #StabilityAI #StableLM3B #LanguageModels https://
bit.ly/3thtyAJ