A floral poster in the style of Kehinde Wiley with an embrace theme and the word “Open Source” in bold text
Sequoia backs Pydantic to expand beyond its open source data-validation framework | TechCrunch
Cohere
“Introducing Takane, an industry-best custom-built Japanese model, developed by Cohere and @Fujitsu_Global.
Updates to Command R Fine-tuning
Cohere just made it way easier for companies to create their own AI language models | VentureBeat
Hugging Face
VoiceRestore – a Hugging Face Space by jadechoghari
Expression Editor – a Hugging Face Space by fffiloni
WebLLM Playground – a Hugging Face Space by cfahlgren1
Model Memory Utility – a Hugging Face Space by hf-accelerate
Molmo 7B D 0924 – a Hugging Face Space by akhaliq
DepthCrafter – a Hugging Face Space by tencent
Open VLM Video Leaderboard – a Hugging Face Space by opencompass
Eagle X5 13B Chat – a Hugging Face Space by NVEagle
Introducing the Open FinLLM Leaderboard
Flux.1-dev Upscaler – a Hugging Face Space by jasperai
Tiny Test Models
Meta/Llama
“Build a researcher-writer multi-agent full-stack app with one CLI command 🔥 create-llama lets you spin up a complete agent template, powered by @llama_index workflows (in Python and Typescript), along with the front-end/backend services. Clone it as a starter for any
“create-llama is the easiest way to get started with multi-agents using @llama_index. v0.2.13 allows you to have a dialogue with the agents. See below creating a blog post about @MetaQuestVR. 🖥️ $ npx create-llama@latest 📄
“Meta AI Llama 3.2 can edit selected parts of the image
Llama 3.2 – a meta-llama Collection
“Meta announced Llama 3.2, the latest iteration of the open source model family Updates include two medium-sized vision LLMs (11B and 90B) and two lightweight, text-only models (1B and 3B) that fit onto mobile devices and are SOTA for many on-device use cases
Llama can now see and run on your device – welcome Llama 3.2
Llama-Vision-11B – a Hugging Face Space by huggingface-projects
Meta Releases Llama 3.2—and Gives Its AI a Voice | WIRED
“Is your LoRA model giving you nonsense output with the Llama chat template? Today @maximelabonne and I discovered a nasty footgun with LoRA fine-tuning and chat templates that have special tokens: you must include the embedding layer and LM head in the trainable parameters! This
Mistral
“20/ @sophiamyang shared a guide on observability with @MistralAI. Proud to see @AgentOpsAI featured alongside our peers.” / X
NVIDIA
nvidia/NVLM-D-72B · Hugging Face
Nvidia stock edges up as industry report predicts ‘unprecedented levels’ of investment in AI data centers
Nvidia just dropped a bombshell: Its new AI model is open, massive, and ready to rival GPT-4 | VentureBeat
https://venturebeat.com/ai/nvidia-just-dropped-a-bombshell-its-new-ai-model-is-open-massive-and-ready-to-rival-gpt-4/
“Nvdia’s NVLM-D 1.0 72B is 🔥 On par with Llama 3.1 405B on Math and Coding. Creative Commons Attribution Non Commercial 4.0 International 🚀 Model Overview: – Multimodal LLM family achieving SOTA on vision-language tasks – Competitive with GPT-4o and open models like Llama https://x.com/rohanpaul_ai/status/1841497232253890937
Leave a Reply