Westlake-AI/SEMA
Switch EMA: A Free Lunch for Better Flatness and Sharpness
This project helps machine learning engineers and researchers improve the performance and speed of their deep neural networks. By applying a simple modification to the Exponential Moving Average (EMA) technique, it takes your trained model parameters and outputs a more robust and efficient model. This is especially useful for those working on computer vision or natural language processing tasks who want to achieve better model generalization.
No commits in the last 6 months.
Use this if you are training deep neural networks for tasks like image classification, object detection, or language modeling and want to enhance your model's stability and generalization without significant extra computational cost.
Not ideal if you are not working with deep learning models or if your primary concern is not model generalization and training efficiency.
Stars
28
Forks
2
Language
Python
License
Apache-2.0
Category
Last pushed
Feb 16, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/Westlake-AI/SEMA"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
jayin92/Skyfall-GS
Skyfall-GS: Synthesizing Immersive 3D Urban Scenes from Satellite Imagery
Tencent-Hunyuan/Hunyuan3D-2
High-Resolution 3D Assets Generation with Large Scale Hunyuan3D Diffusion Models.
ActiveVisionLab/gaussctrl
[ECCV 2024] GaussCtrl: Multi-View Consistent Text-Driven 3D Gaussian Splatting Editing
caiyuanhao1998/Open-DiffusionGS
Baking Gaussian Splatting into Diffusion Denoiser for Fast and Scalable Single-stage Image-to-3D...
deepseek-ai/DreamCraft3D
[ICLR 2024] Official implementation of DreamCraft3D: Hierarchical 3D Generation with...