https://yanndupis.github.io/index.html Practical experiments and lessons across AI training, evaluation, inference, and production. en quarto-1.1.251 Fri, 17 Jul 2026 07:00:00 GMT Yann Dupis https://yanndupis.github.io/posts/tool-call-streaming/index.html tool_calls event. ]]> local models tool calling inference https://yanndupis.github.io/posts/tool-call-streaming/index.html Fri, 17 Jul 2026 07:00:00 GMT Yann Dupis https://yanndupis.github.io/posts/moose-hf-inference/moose_hf_blog.html Sentiment Analysis on Encrypted Data with Homomorphic Encryption co-written by Zama and HuggingFace. Zama has created an excellent encrypted machine learning library, Concrete-ML, based on fully homomorphic encryption (FHE). Concrete-ML enables data scientists to easily turn their machine learning models into an homomorphic equivalent in order to perform inference on encrypted data. In the blog post, the authors demonstrate how you can easily perform sentiment analysis on encrypted data with this library. As you can imagine, sometimes you will need to perform sentiment analysis on text containing sensitive information. With FHE, the data always remains encrypted during computation, which enables data scientists to provide a machine learning service to a user while maintaining data confidentiality. ]]> deep learning mpc cryptography https://yanndupis.github.io/posts/moose-hf-inference/moose_hf_blog.html Mon, 09 Jan 2023 08:00:00 GMT Yann Dupis https://yanndupis.github.io/posts/perceiver-io-experiment/perceiver_io_from_scratch.html Pytorch Image Models (timm) library. As of now it contains more than 50 different types of architectures translating into 600 models which have different trade-offs in terms of accuracy and inference time, depending on the task. For NLP, you can use the popular transformer library from HuggingFace which offers almost 80k models based on about 140 different types of architectures. ]]> deep learning code from scratch https://yanndupis.github.io/posts/perceiver-io-experiment/perceiver_io_from_scratch.html Mon, 31 Oct 2022 07:00:00 GMT