MCP App StoreLe meilleur de l'écosystème IA et MCP, sélectionné chaque jour.
Sources
Hugging Face and Microsoft are strengthening their partnership to better integrate open-source AI models into developer workflows and cloud infrastructure. This collaboration aims to democratize AI access and accelerate the deployment of open-science models.
Hugging Face announces official support for AMD Instinct MI300 GPUs, enhancing hardware diversity for AI workloads. This allows developers to leverage AMD's high-performance compute for training and inference using HF libraries.
Hugging Face partners with Dell to enable on-premise AI deployment via the Dell Enterprise Hub. This integration simplifies the process of running open-source models on local infrastructure for better data control and security.
A deep dive into KV cache quantization techniques to reduce memory overhead and enable significantly longer sequence generation in LLMs. Crucial for developers optimizing high-throughput inference and large-context window applications.
Hugging Face and LangChain have released a dedicated partner package to streamline the integration of Hugging Face models and datasets into LangChain workflows. This simplifies deploying state-of-the-art AI agents and LLM applications.
Google introduces PaliGemma, an open vision-language model designed for fine-tuning on downstream tasks like image captioning and visual QA. It provides a powerful base for developers to build specialized multimodal applications.
Hugging Face introduces Transformers Agents 2.0, enhancing the ability of LLMs to interact with tools and environments. This update provides a more robust framework for building agentic workflows using the Transformers library.
A technical guide on optimizing RAG applications using Intel Gaudi 2 and Xeon hardware. It focuses on cost-efficiency and performance scaling for enterprise-grade retrieval augmented generation pipelines.
Explores methods for improving the consistency of LLM outputs using structured generation techniques. Essential for developers building reliable agents that require strict data formats.
Introducing StarCoder2-Instruct, a transparently aligned model optimized for code generation. It provides a permissive license and high-quality performance for AI-assisted development workflows.
Meta has released Llama 3, significantly advancing the capabilities of open-weights LLMs. This release provides a powerful alternative for developers building local or self-hosted AI applications.