Brett Young @BrettYoung3d

Explore fine-tuning the new Llama 3 with TorchTune and HuggingFace. Find out which tool enhances your ML projects best. Read here → bit.ly/44g62T6

Developers, Developers, Developers — Satya when GPT5 launches

Geoffrey Hinton says AI models have intuition, creativity and the ability to see analogies that people cannot see

📚 Discover how to optimize GPU use with DeepSpeed Zero and HuggingFace Trainer. Our guide simplifies maximizing model training efficiency. Dive in now→ wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

🔎 Discover the journey of a first-time Kaggle participant and the lessons learned beyond coding. This concise blog offers insights into problem-solving and community engagement in data science. Dive in now → wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

🔍 Exploring the potentials of DeepSpeed Zero with the HuggingFace Trainer for enhanced GPU efficiency is key. Dive into our comprehensive guide that refines large model training methods. Read here: wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

🍽️ Elevate Your Dining with Our AI-Interactive Menu! Say goodbye to static menus. Leveraging RAG, @llama_index, & W&B Weave, explore our blog to see how we're shaping the future of restaurant menus. Read more here: wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

🌋 Looking to Fine-Tune LLaVA on Your Custom Dataset? Our tutorial makes it simple! Learn to adapt the LLaVA model with GPT-4 for your datasets. It's the next level of AI integration, combining text & visuals seamlessly. Read now🔗 wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

Speculative execution for LLMs is an excellent inference-time optimization. It hinges on the following unintuitive observation: forwarding an LLM on a single input token takes about as much time as forwarding an LLM on K input tokens in a batch (for larger K than you might…

Georgi Gerganov @ggerganov

8 months ago

Speculative execution for LLMs is an excellent inference-time optimization. It hinges on the following unintuitive observation: forwarding an LLM on a single input token takes about as much time as forwarding an LLM on K input tokens in a batch (for larger K than you might…

40 270 2K 1.1M 709

Download Video

New in LLM: Mixtral 8x7B put to the test with MMLU & W&B. In this article, we run Mixtral 8x7B through its paces with the MMLU dataset and W&B. Find out what sets this model apart now👉 wandb.ai/byyoung3/ml-ne…

AI meets Legal! 📜⚖️ AI is revolutionizing the legal sector, promising enhanced efficiency and accessibility. Discover how @Azure, @OpenAI, and W&B are transforming contract analysis. Read now👉 wandb.ai/byyoung3/ml-ne… (@BrettYoung3d)

🤔 Ever wondered how to run Mistral 7B on your M1 Mac? Learn through our tutorial how to deploy this powerful AI model, including on M series Macs. Dive into our step-by-step guide and discover the potential of AI on your desktop! 🔗 bit.ly/4aizGJV (@BrettYoung3d)

👨‍💻 Mistral 7B or Zephyr 7B: Which AI model excels at coding? Find out in our latest blog! We put them to the test on the HumanEval benchmark. Dive in now🔗 bit.ly/40T3CI0 (@BrettYoung3d)

@sama Microsoft’s board right now

Today I resigned from my position as CEO of Cruise. (1/5)

@jasonkwon @sama

Ever thought of stretching LLMs beyond 4096 tokens? 🤔 Delve into our journey with Llama 2, now scaled to 32k tokens with LongLora's capabilities. Advancements in long content processing await! 🔗 bit.ly/45ZJVze (@BrettYoung3d)

Are you fine-tuning Mistral 7B? 👀 Here is a tutorial guide on fine-tuning Mistral7B on Python code with a single GPU! Learn how to push the boundaries of both the performance & efficiency of models here 🔗 bit.ly/3RVDv0C

💡Say hello to Falcon 180B! @huggingface presents the most expansive open-source language model to date, with 180 billion parameters & 3.5 trillion-token dataset. Developed by TII, it's a game-changer in NLP. Learn more about this critical tool 🔗 bit.ly/3LhvwHd

The dynamics of adopting new libraries are changing 🪄 Keras Core now provides seamless interaction between Torch, Tensorflow, & JAX so that engineers no longer need to be confined to a single one. Read everything you need to know here bit.ly/3rsNDmm (@BrettYoung3d)