ML Research Hub – Telegram
ML Research Hub
32.7K subscribers
4.01K photos
229 videos
23 files
4.32K links
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.

Admin: @HusseinSheikho || @Hussein_Sheikho
Download Telegram
🦙 Fintuning Llama 3 using ORPO.

A quick guide on how to set up your new Llama 3 8B with ORPO .

I hope you will enjoy!

🤗 Model : https://huggingface.co/mlabonne/OrpoLlama-3-8B

💻 Colab : https://colab.research.google.com/drive/1eHNWg9gnaXErdAa8_mcvjMupbSS6rDvi?usp=sharing

📝 Article : https://huggingface.co/blog/mlabonne/orpo-llama-3

https://news.1rj.ru/str/DataScienceT
Please open Telegram to view this post
VIEW IN TELEGRAM
🦾 🦏 Power of matplotlib

This beauty can be made using matplotlib . This is a visualization of an engraving by the German artist Albrecht Dürer, depicting an Indian rhinoceros, as the artist imagined it from the denoscriptions and drawings available to him in 1515.

Want to learn the same thing: here's a cool free book: " Scientific Visualization: Python + Matplotlib "

The sources of the book with code examples are here .

Poster
Book
Code from the book

https://news.1rj.ru/str/DataScienceT
Please open Telegram to view this post
VIEW IN TELEGRAM
👍6
This media is not supported in your browser
VIEW IN TELEGRAM
🛞 6Img-to-3D driving scenarios 🛞

👮‍♀️ EPFL (+ Continental) unveils 6Img-to-3D, novel transformer-based encoder-renderer method to create 3D onbounded outdoor driving scenarios with only six pics

🥺 Review: https://shorturl.at/dZ018

🤨 Paper: arxiv.org/pdf/2404.12378.pdf

👉 Project: 6img-to-3d.github.io/

👉 Code: github.com/continental/6Img-to-3D

https://news.1rj.ru/str/DataScienceT
Please open Telegram to view this post
VIEW IN TELEGRAM
👍4
📑 Big data and artificial intelligence in cancer research

📕 Journal:  Trends in Cancer (🔥 I.F.= 18.4)
🗓 Publish year: 2023

📱 Authors: Xifeng Wu, Wenyuan Li, Huakang Tu
🏢 University: Zhejiang University School of Medicine, China

📎 Study the paper

https://news.1rj.ru/str/DataScienceT ⚙️
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍2
🐱 Moving Object Segmentation:All You Need Is SAM (and Flow)

SAM + Optical Flow = FlowSAM

FlowSAM is a new tool for detecting and segmenting moving objects in video, which significantly outperforms all previous models , both for a single object and for multiple objects 🔥

Project page: https://www.robots.ox.ac.uk/~vgg/research/flowsam/

Code: https://github.com/video2game/video2game

Paper: https://arxiv.org/abs/2404.12389

Data: https://drive.google.com/drive/folders/1tmDq_vG_BvY5po40Ux5OBds1avUM_CbR

https://news.1rj.ru/str/DataScienceT ⚙️
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍6
📁 Machine Learning and Health Science Research: Tutorial

📕 Journal: Journal of Medical Internet Research (I.F.= 7.4)
🗓 Publish year: 2024

📱 Authors: Hunyong Cho, Jane She, Daniel De Marchi, ...
🏢 University: University of North Carolina at Chapel Hill, United States

📎 Study the paper
Please open Telegram to view this post
VIEW IN TELEGRAM
👍1
This media is not supported in your browser
VIEW IN TELEGRAM
🗣 OpenVoice V2 - a Text-to-Speech model that can clone any voice and speak in any language.

OpenVoice V2 is a new version of the open text-to-speech model that allows you to clone any voice and generate speech in various languages.

Github: https://github.com/myshell-ai/OpenVoice/tree/main
Usage: https://github.com/myshell-ai/OpenVoice/blob/main/docs/USAGE.md
Please open Telegram to view this post
VIEW IN TELEGRAM
👍51
🖼 HiDiffusion: Unlocking High-Resolution Creativity and Efficiency in Low-Resolution Trained Diffusion Models 🦊

A new training-free method that improves the performance of pre-trained diffusion models.

It can be integrated into diffusion pipelines by adding just one line of code!

pip3 install hidiffusion


page : https://hidiffusion.github.io
paper : https://arxiv.org/abs/2311.17528
code : https://github.com/megvii-research/HiDiffusion
colab : https://colab.research.google.com/drive/1EiBn9lSnPZTU4cikRRaBBexs429M-qty?usp=sharing

https://news.1rj.ru/str/DataScienceT ⚙️
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍61
🔥 FlowMap: High-Quality Camera Poses, Intrinsics, and Depth via Gradient Descent

Massachusetts Institute of Technology introduced FlowMap.

This is a new comprehensive differentiable method for reconstructing a 3D scene , which allows you to accurately specify camera angles, motion characteristics and video depth for each frame.

FlowMap allows you to create realistic 360° views.

Github: https://github.com/dcharatan/flowmap
Paper: https://arxiv.org/abs/2404.15259
Dataset: https://drive.google.com/drive/folders/1PqByQSfzyLjfdZZDwn6RXIECso7WB9IY

https://news.1rj.ru/str/DataScienceT ⚙️
Please open Telegram to view this post
VIEW IN TELEGRAM
👍52
OpenBioLLM-Llama3-70B and 8B: Open-source Large Language Models in Medical Domain

OpenBioLLM-Llama3-70B and 8B: the most effective and affordable Lms in the medical field today! 🩺 💊 🧬

Outperforms industry giants like GPT-4, Gemini, Meditron-70B, Med-PaLM-1 and Med-PaLM-2 in the biomedical field. 🏥 📈 ⭐️

The OpenBioLLM-70B reaches SOTA and is a new achievement for models of this size.
The OpenBioLLM-8B even outperforms GPT-3.5, Gemini and Meditron-70B! 🚀

- 70B : https://huggingface.co/aaditya/OpenBioLLM-Llama3-70B

- 8B : https://huggingface.co/aaditya/OpenBioLLM-Llama3-8B

- Medical Leaderboard : https://huggingface.co/spaces/openlifescienceai/open_medical_llm_leaderboard
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍21
📃Graph Machine Learning in the Era of Large Language Models (LLMs)

🗓 Publish year: 2023

🧑‍💻Authors: Wenqi Fan, Shijie Wang, Jiani Huang, Zhikai Chen, Yu Song, Wenzhuo Tang, Haitao Mao, Hui Liu, Xiaorui Liu, Dawei Yin, Qing Li

🏢Universities: The Hong Kong Polytechnic University,Michigan State University, North Carolina State University

📎  Study the paper
👍3
Acquired global fresh database Please send me a sample Scammer please don't waste your time @yuefu666
⭐️ The PyTorch team is developing a library for learning LLM called torch titan.

Today the library has become publicly available on GitHub, but it is still in a pre-release state and is actively being developed.

- Library link: https://github.com/pytorch/torchtitan

- Tutorial on working with torch titan: https://www.youtube.com/watch?v=ee5DOEqD35I

The library was created for preliminary training of models, and for fine tuning PyTorch has another library, torchtune:
https://github.com/pytorch/torchtune

https://news.1rj.ru/str/DataScienceT ⚙️
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍6❤‍🔥31🏆1