BADAOUI Abdennacer

About

BADAOUI Abdennacer works in ML & data science and AI engineering, based in Paris, France (HuggingFace). ML Engineer @huggingface. Public projects on GitHub include D3PMs (A small project that uses Discrete Denoising Diffusion Probabilistic Models (D3PMs), a generative model for…), Reducing_the_Transformer_Architecture_to_a_Minimum (An implementation of the techniques introduced in the paper "Reducing the Transformer Architecture to a…) and Connected-objects-detection (SÉCURITÉ ET DÉFENSE ÉCOUTE ET HACKING: Machine Learning appliqué à la détection et à la reconnaissance…), mostly written in Jupyter, Python and Java.

Projects 6

• 

  D3PMs

  A small project that uses Discrete Denoising Diffusion Probabilistic Models (D3PMs), a generative model for discrete data that builds upon the success of Denoising Diffusion Probabilistic Models (DDPMs) in continuous spaces.

  github.com
• 

  Reducing_the_Transformer_Architecture_to_a_Minimum

  An implementation of the techniques introduced in the paper "Reducing the Transformer Architecture to a Minimum"

  github.com
• 

  Connected-objects-detection

  SÉCURITÉ ET DÉFENSE ÉCOUTE ET HACKING: Machine Learning appliqué à la détection et à la reconnaissance d'objets connectés ZigBee

  github.com
• 

  Predict-Product-Categories-and-Brands-Using-CLIP-Embeddings

  In this project, we aim to classify product categories (based on the Google taxonomy) and brands using fine-tuned CLIP embeddings. Our primary focus will be on aggregating textual and image embeddings to achieve a more representative depiction of the products, leading to enhanced performance.

  github.com
• 

  Stable_Diffusion_From_Scratch

  Stable Diffusion from scratch

  github.com
• 

  Sort Your Prompts

  Sorting prompts before batching speeds up MoE decoding

  github.com