Transformer on Noureddine RAMDIhttps://ramdi.fr/tags/transformer/Recent content in Transformer on Noureddine RAMDIHugoenSat, 23 May 2026 20:41:27 +0000Tracing deep learning step-by-step in Excel: a hands-on guide to ai-by-hand-excelhttps://ramdi.fr/github-stars/tracing-deep-learning-step-by-step-in-excel-a-hands-on-guide-to-ai-by-hand-excel/Sat, 23 May 2026 20:41:14 +0000https://ramdi.fr/github-stars/tracing-deep-learning-step-by-step-in-excel-a-hands-on-guide-to-ai-by-hand-excel/Explore how ai-by-hand-excel implements deep learning architectures like Transformers entirely in Excel formulas, exposing the math behind AI step-by-step without code.daVinci-MagiHuman: Simplifying multimodal video and audio generation with a single-stream transformerhttps://ramdi.fr/github-stars/davinci-magihuman-simplifying-multimodal-video-and-audio-generation-with-a-single-stream-transformer/Mon, 04 May 2026 10:23:02 +0000https://ramdi.fr/github-stars/davinci-magihuman-simplifying-multimodal-video-and-audio-generation-with-a-single-stream-transformer/daVinci-MagiHuman uses a 15B-parameter single-stream transformer with a sandwich architecture to generate video and audio from text, achieving competitive quality and fast inference on a single H100 GPU.MicroGPT-C: Coordinating tiny GPT-2 models in C for edge logical reasoninghttps://ramdi.fr/github-stars/microgpt-c-coordinating-tiny-gpt-2-models-in-c-for-edge-logical-reasoning/Mon, 04 May 2026 10:23:02 +0000https://ramdi.fr/github-stars/microgpt-c-coordinating-tiny-gpt-2-models-in-c-for-edge-logical-reasoning/MicroGPT-C uses a deterministic C scaffold to coordinate tiny GPT-2 models, achieving 90%+ accuracy on logic games with 8x memory compression and infinite sequence lengths.Understanding LLM internals: a hands-on guide to transformers and attention mathhttps://ramdi.fr/github-stars/understanding-llm-internals-a-hands-on-guide-to-transformers-and-attention-math/Mon, 04 May 2026 10:23:02 +0000https://ramdi.fr/github-stars/understanding-llm-internals-a-hands-on-guide-to-transformers-and-attention-math/A curated repo breaking down large language model internals with numeric attention math, tokenization, and transformer architecture, targeting engineers who want to understand LLMs under the hood.Inside llm-madness: a lightweight GPT transformer training pipeline with built-in visualizationhttps://ramdi.fr/github-stars/inside-llm-madness-a-lightweight-gpt-transformer-training-pipeline-with-built-in-visualization/Mon, 04 May 2026 10:23:01 +0000https://ramdi.fr/github-stars/inside-llm-madness-a-lightweight-gpt-transformer-training-pipeline-with-built-in-visualization/llm-madness offers a Python-built GPT-style transformer training pipeline with tokenizer training, memory-mapped datasets, and a unique web UI for per-layer attention inspection and loss visualization.NOVA3R: Non-pixel-aligned visual transformer for amodal 3D reconstruction from unposed multi-view imageshttps://ramdi.fr/github-stars/nova3r-non-pixel-aligned-visual-transformer-for-amodal-3d-reconstruction-from-unposed-multi-view-images/Mon, 04 May 2026 10:23:01 +0000https://ramdi.fr/github-stars/nova3r-non-pixel-aligned-visual-transformer-for-amodal-3d-reconstruction-from-unposed-multi-view-images/NOVA3R implements a non-pixel-aligned visual transformer for amodal 3D reconstruction from unposed multi-view images, recovering occluded geometry with physical plausibility.