12 articles
CLIPSeg combines CLIP with segmentation for zero-shot image segmentation. The model uses text descriptions to identify and isolate specific objects or regions in images without requiring prior training or annotations.
Hugging Face introduces Model Cards, standardized documentation sheets for AI models covering architecture, training data, performance metrics, limitations, and use cases. This tool enhances transparency and simplifies model evaluation before deployment.
OpenAI introduces Point-E, a system generating 3D point clouds from complex text prompts. The model uses a two-stage process: intermediate image generation followed by 3D geometry conversion. Point-E produces results 60x faster than existing 3D diffusion methods.
OpenAI announces a new embedding model that is significantly more capable, cost effective, and simpler to use.
Hugging Face releases the second issue of its Ethics and Society Newsletter, focusing on biases in machine learning. The article examines bias sources, societal impacts, and mitigation approaches in AI systems.
Complete guide to audio datasets for AI: sources, formats, preparation and best practices. Covers annotation, cleaning, augmentation and evaluation for training robust models.
Habana Gaudi2 outperforms Nvidia A100 80GB in training and inference speed across multiple LLM models. Comparative benchmark showing performance gains on Llama 2, Mistral and other architectures.
Hugging Face releases official Elixir integration, enabling developers to access GPT-2, Stable Diffusion and other models through the platform. First step toward ML ecosystem in Elixir.
Hugging Face publishes an educational illustration of RLHF (Reinforcement Learning from Human Feedback) process. The article details how language models are fine-tuned via human feedback and reinforcement optimization to improve alignment with user preferences.
Hugging Face publishes a guide on deep learning applied to proteins. The article covers model architectures, datasets, and training techniques for predicting protein structures and properties.
Hugging Face releases a guide for integrating Stable Diffusion with Core ML on Apple Silicon. The solution enables local image generation on Mac without relying on remote servers, optimizing native hardware performance.
Hugging Face adds probabilistic time series forecasting to Transformers library. New feature generates probability distributions for predictions beyond point estimates. Includes pre-trained models and fine-tuning support.