Announcing Concrete ML v0.4

Zama has released a new version of Concrete-ML, featuring client/server APIs for deployment, new machine-learning models, improved processing speed, and support for Quantization Aware Training (QAT). The client/server APIs enable key generation, data encryption, model execution on untrusted servers, and result decryption on clients, facilitating production deployment of models. The release expands available machine-learning models, adding regressors like Lasso, Ridge, ElasticNet, and tree-based models like DecisionTree and RandomForest. An important advancement is the introduction of QAT, which enhances model accuracy by optimizing weights under low bit-width constraints, making models more effective compared to previous versions. This update includes integration of QAT into built-in neural network models and the ability for users to import custom quantized models, with Brevitas being used for quantization on datasets like MNIST. Future efforts will focus on tackling complex tasks using QAT and extended precision in computation.

No tracked trend matches for this post yet.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.