Matteo Cappelloni’s Post

Lost in Hyperspace has been Pwned! . . . Let’s proceed with the steps: * Load the Embeddings: First, we'll load the provided embeddings file. * Apply Dimensionality Reduction: We'll apply t-SNE (or PCA) to reduce the dimensionality for visualization. * Sort and Extract Tokens: After dimensionality reduction, we'll sort and extract tokens, then concatenate them to form the flag.

Today we're releasing SmolVLM, a small 2 billion parameters Vision+Language Model (VLM) built for on-device/in-browser inference which outperforms all models at similar GPU RAM usage and tokens throughputs SmolVLM can be fine-tuned on a Google collab and be run on a laptop... or process millions of documents with a consumer GPU! Work led by Andrés Marafioti and the SmolLM/multimodal teams at Hugging Face As always everything is open-source! Some links: demo: https://lnkd.in/e4HT-UvB blog post: https://lnkd.in/eg_v43WJ dataset: https://lnkd.in/eKFCa969 finetuning notebook: https://lnkd.in/eUwmBiTS

💡 Want to make your ComfyUI workflow shareable while running it on a powerful GPU? 👉 Check out Het Trivedi and Philip Kiely's guide on how to deploy ComfyUI pipelines behind an API endpoint in minutes: https://lnkd.in/ehFX4FGa 🏎️ Run fast inference, and use your workflow in any application. If you're using custom nodes or model checkpoints, Het Trivedi and Rachel Rapp have a guide for that, too: https://lnkd.in/ejDJMv7Q

“Object tracking” is an exciting technology that promises to significantly transform object-oriented learning, work, and communication. I have been working for years on the concept of “Object-Oriented Transformation” to keep pace with this development. According to the vision I am working on, there will be a need to reformulate business activities and functions, in a way that integrates language, work activities and senses, to keep pace with a hybrid and immersive environment, by tracking objects. This vision provides integrated headings for business functions in an object-oriented manner, It helps in organizing and coordinating roles, efforts, and setting priorities in a hybrid (physical-digital) world.. These headings focus on the following: 1- What is the relationship of an object to its surroundings? 2-What is the ground on which the object stands? 3-How do Objects cooperate with each other? 4-How do Objects cooperate with each other over time. 5- What are the states that an Objects can go through? 6- How will the work or object be modeled logically? 7- How the components of the product (object) will be physically connected. 8-What is the most appropriate method (in terms of form and mechanism) to deliver the product to the customer? Apple Forbes LVMH Richemont Bvlgari Gucci BMW Group AUDI AG Toyota Motor Corporation Ford Motor Company School of International Futures (SOIF) BCG X

This is my Stable Diffusion Lab repo using Google Colab, created specifically to make it easier for anyone to learn and use the existing features. This repo will be continuously updated over time so that everyone can try out new things in the development of stable diffusion using Google Colab. I put it in Google Colab because there is a free GPU available that can help speed up inference. Link the Repo: https://lnkd.in/gGiaQfRd IMPORTANT: Please use it wisely and responsibly.

Your terminal reimagined IDE-style text input/navigation  Block-based command grouping Ability to save and share commands Warp AI can generate commands from normal text Customise keybindings and launch configs Built-in themes + support for custom ones

For any one working with text-to-speech (coqui TTS) and wondering how all the combinations of models + vocoder sound , have compiled them here for easy reference with some recommendations. With multiple models, datasets and vocoders to choose from, i wanted to get a baseline of which combination sounds better for my use case so that can serve as a base model for further fine tuning. Hence this compilation !! Hope it helps someone who is in similar journey or needs. Github link: https://lnkd.in/gjF5xNiy #tts #texttospeech #coqui

A couple of days back, I stumbled upon rabbit inc., a totally personalized, AI-based OS fitted inside a pocket-sized device. What's cool about it? It's based on an LLM interacting with a LAM (Large Action Model), a cutting-edge foundation model that fully understands our needs when interacting with digital services. Here's a rundown of what this system can offer: - Not only can you interact with it thanks to the LLM, but it also can interact with external sites and apps to produce real outcomes (think of ordering at Uber or buying something at Amazon...) - Not only does it get the big picture, but it's also all about the little details. You can train it to recognize and reproduce specific actions or sequences so it can replicate them later on (think of routines, like turning on the AC, ordering at Uber, and messaging your pals for a Friday night at the same time, all tied to your voice command). Imagine a world where your smartphone's on the same wavelength as your best bud. Can't wait to explore its potential!

Interesting overview of test-time compute scaling for higher accuracy from smaller models (ex. Lllama 3B outperforms 70B on accuracy when allocated additional compute); TLDR: introduce high diversity at the front end and use a variety of search / verification tools to optimize the final output. This study looks at the MATH-500 evaluation set, but the technique would get really interesting when coupled with a code interpreter and parallel unit test generation model.

Does YouTube's datacenter have storage issues? Probably not, but they now have a bit less space after uploading my latest F5 voice cloning tutorial 😆. Want your own high-quality voice clone running locally on your PC with just 10 seconds of audio input? Yes, it's possible with F5! Check out my step-by-step tutorial with audio samples on my "Thorsten-Voice" YouTube channel. ➡️ https://lnkd.in/eUFb6erx #AIVoice #TextToSpeech #Tutorial #VoiceCloning