Did you watch Eurovision 🤡 last night?

MLX Guru has been updated to version 0.16.1 Awni Hannun Angelos Katharopoulos

🚀 MLX Guru: not perfect but can speed up navigating the framework and most of the time offers good code. As you get more experienced with MLX you can prompt it around occasional coding suggestion fails Awni Hannun Angelos Katharopoulos

Thermo Therapist with CoolProp is being released in the GPT store for my Engineering Thermo students. It explains concepts, uses a custom API to access CoolProp, solves problems & will convince you that you should ❤️ Thermo. Not responsible for any Transference issues 😂 OpenAI

MLX Whisper package now works with Distil-Whisper (and other Transformers compatible Whisper models). distil-large-v3 runs 40X faster than realtime on my M1 Max (transcribes 12 minutes in 18 seconds) pip install -U mlx-whisper Example:

Using MLX Whisper with distil-whisper-large-v3 I was able to transcribe in 90 seconds and with basically perfect accuracy a 40 minute video of the lecture “Beyond Language: Applying MLX Transformers to Engineering Physics” that I had recorded for my students 🚀🚀🚀 Awni Hannun

Watch a video of “Thermo Therapist with CoolProp” in action doing CoolProp calculations and explaining Thermodynamics 🚀 OpenAI vimeo.com/997825122/9414…

I had to try it 😄😄: diffusionkit-cli --prompt "A coffee cup with the print 'I code with MLX and triple espresso'" --output-path image.png --model-version FLUX.1-schnell --steps 6”. 20 secs on M2U. Awni Hannun

Nice update coming to MLX LM: better scaling for much longer prompts / generations. Processing a 10,000 token prompt with 4-bit LLama 3.1 using only ~5GB:

Απίστευτο αλλά αληθινό 🤣🤣🤣 Kudos to OpenAI

pip install -U mlx

🫁🌊 Collaborating with Josué Sznitman on this Annual Review was fun & rewarding. It is finally live in early release Annual Reviews : Multiscale Modeling of Respiratory Transport Phenomena and Intersubject Variability | Annual Reviews - go.shr.lc/3Z5KXKj

Made a fast implementation of Conway's Game of Life accelerated with MLX custom kernels. 30 lines of code for the whole function and it runs super fast accelerated on the GPU. Example: gist.github.com/awni/fde217c67… Random 512x512 grid animated for fun:

If you’re applying Transformers in Engineering consider submitting your work to the Special Issue of Results in Engineering dedicated to the application of Transformers in Engineering. Co-Edited by Awni Hannun Angelos Katharopoulos, Alessio Alexiadis & myself. Submit: tinyurl.com/ys59ucz7

GPT-o1-preview is spot on ✅ (given the hint). Definitely a significant improvement in reasoning. This used to fail with GPT-4 and produced a rather convoluted answer with GPT-4o. Sam Altman OpenAI

Starting today we initiate a series of science podcasts 🎧 on X showcasing some of the most exciting research output of UCY-CompSci UCY-RespiHub & collaborators. Thanks to Google for making this possible. These podcasts are themselves a testimony of the real value of AI 🚀🚀

The second Podcast 🎧 in our series dives into our efforts UCY-CompSci with collaborators to create a computationally efficient model of the deep lung 🫁 by taking some clever shortcuts: The cited article is doi.org/10.1016/j.jaer… #AerosolScience #InhalationTherapies #LungHealth