Models & Labs
NVIDIA and Hugging Face Enhance Diffusion Model Training
Hugging Face Bloghigh confidence
Why it matters
- →Simplifies the training process for diffusion models by eliminating the need for checkpoint conversion.
- →Supports scalable training configurations, making it easier to train large models.
- →Enhances flexibility with both full and parameter-efficient fine-tuning options.
NVIDIA and Hugging Face have collaborated to enhance the training of diffusion models with the NeMo Automodel library. This integration allows for seamless training of models from the Hugging Face Hub without the need for checkpoint conversion or model rewrites. It supports scalable training configurations and both full and parameter-efficient fine-tuning. This advancement simplifies the process for developers working with diffusion models, making it easier to train and deploy these models at scale.
Read originalMore from Hugging Face Blog
© Hugging Face BlogModels & Labsmodels
NVIDIA Nemotron 3 Embed Tops RTEB Leaderboard
NVIDIA's Nemotron 3 Embed models have set a new benchmark in retrieval quality, with the 8B model ranking #1 on the RTEB leaderboard. This collection of embedding models is designed for production-scale retrieval tasks, offering open weights and datasets for customization. The models support multilingual and code retrieval, and are optimized for high-throughput deployment with NVIDIA's NVFP4 technology. This release provides developers with powerful tools for efficient and accurate retrieval, enhancing capabilities in agentic retrieval and reducing operational costs.
Hugging Face Blog
© Hugging Face BlogMore in Models & Labs
Models & Labsmodels
llama.cpp b10057 release focuses on SYCL fixes
The b10057 release of llama.cpp targets critical SYCL-related issues, enhancing the stability of its computation kernels. A significant fix addresses a row calculation error when K_QUANTS_PER_ITERATION is set to 1, ensuring more accurate results. Additionally, the update corrects the processing of reordered q5_k kernels, which is crucial for maintaining performance integrity. These improvements, contributed by Intel's Todd Malsbary, are designed to bolster the accuracy and efficiency of SYCL operations. While no new features are introduced, the release strengthens the existing framework, providing developers with a more reliable environment for SYCL-based computations.
llama.cpp Releases
Models & Labsmodels
llama.cpp b10063 Release Expands Platform Support
The latest b10063 release of llama.cpp continues its trend of broadening platform compatibility, now including support for Vulkan on Ubuntu and Windows, as well as ROCm 7.2 on Ubuntu. This update ensures that developers working across diverse hardware configurations can leverage llama.cpp's capabilities more effectively. Notably, the release maintains its focus on providing robust support for both CPU and GPU environments, including CUDA and OpenVINO. While no groundbreaking features are introduced, the expansion of supported platforms signifies llama.cpp's commitment to being a versatile tool for AI developers.
