Unlock the Potential of AI
SqueezeBits is building an ecosystem for designing, training, and deploying highly compressed Neural Networks.
Our Mission
Faster
From research to real life application
Lighter
The smaller size, the broader applications
Cost Effective
Equal performance at a low cost
Safer
Data only stays where it came from
How we can help
Model Profiling
Model profiling is the most critical first step in model optimization.
Identifying bottlenecks that slow down the model and determining the potential performance improvement through optimization are the key aspects.
Reducing Latency & Size
SqueezeBits optimize the model based on a deep understanding of the model architecture and target hardware.
Based on quantization, SqueezeBits select the most suitable optimization method that fits the application and environment.
Hardware specific optimization
SqueezeBits offers lightweighting services that are optimized for your hardware's characteristics.
Migrate from the server to the edge
By migrating AI computing from the cloud to the edge device, you can reduce the server operating cost.
Why SqueezeBits?
Optimizing an AI model strongly requires deep understanding of the hardware.
SqueezeBits maintain technical expertise on lightweight the models on following devices:
CPU, GPU, Mobile AP, Micro Controller, NPU
Applications
-
Vision
Image classification, Segmentation, Object detection, Face recognition, Pose estimation and more
-
NLP & Speech
Keyword spotting, Question answering, Noise suppression, Translation, Speech recognition and more
-
Generative AI
Stable diffusion on a mobile environment by implementing quantization, pruning, and knowledge distillation and more
AI Compression Toolkit
OwLite is the easiest AI compression toolkit, simplifies the process of applying advanced AI compression techniques to pre-existing machine learning models
Lightweight your AI model with us
SqueezeBits excels in the optimization of deep learning models using state-of-the-art model compression techniques.