Fine-tune LLMs 30x faster! With Daniel Han (Unsloth AI)

[@TheAIEpiphany]

Daniel Han from Unsloth AI joined us on a bug hunting tour of DeepMind's Gemma model! :) Also, if you need some GPUs check out Hyperstack: console.hyperstack.cloud/?Influencers&Aleksa+Gordi%C4%87 who are sponsoring this video! :)

[@TheAIEpiphany]

This talk was much more technical - let me know if it was too much :')

[@danielhanchen]

Thanks for inviting me!! Loved talking about bug hunting esp for Gemma, Llama, Phi-3 and all about Unsloth!!

[@b0nce]

Man you're the best Love your work Thanks for all the things you do

[@robertjalanda]

Have been following Unsloth since the beginning (well since January) and it truly has revolutionized finetuning/training for the GPU poor! So happy to see them get recognition

[@avishvijayaraghavan]

Great chat, Daniel's killing it

[@Jay-wx6jt]

Save this guy. He is a gem. Thank you Daniel for your contribution 🎉

[@Pingu_astrocat21]

Thanks to Daniel! Unsloth has mad fine-tuning so much easier to use :)

[@bigglyguy8429]

I still don't understand it. Is there a GUI for it?

[@alealejandroooooo]

Awesome talk!

[@TimTruth]

Genius- super interesting

[@DevasheeshMishra]

Need more such videos 😊

[@dinorossi6611]

Aleksa, interesantan ti je kanal. Vidim da si self taught machine language programmer. Josh sam nov u svemu tome ali imam neke fenomenalne ideje....

[@baner3744]

Damn casting was the contributor inspite of DSA algorithm masters hired by big tech , coding like this does not happen in Government/ Non-profit where less paid programmers toil

[@kimchi_taco]

Glad to watch! RoPE, emb * sqrt(fan_in), RMSNorm need f32!!! 32:00 Alexa, awesome question!