Video wird geladen...
Video konnte nicht geladen werden
I was able to run Meta’s Llama 2 7B parameters locally without internet connection on my 15 Pro Max It was a bit hot and a dumb cuz small LLM as a beta app (LLM Farm) Imagine it was a iOS centric and optimized one for Apple uses tho 🤌🏻
17,614 Aufrufe • vor 2 Jahren •via X (Twitter)
8 Kommentare

AppleLeakervor 2 Jahren
Wow. Could you please share how you were able to run this on-device?

ShrimpApplePro 🍤 ずっと真夜中でいいのにvor 2 Jahren
LLM Farm (Testflight) You can use the sample models or trying others on hugginface I think you can get a GPT 2 as well I would recommend run it on a M1 machine or A17 Pro cuz holy hell it eats RAM and hot af on first run 💀

ShrimpApplePro 🍤 ずっと真夜中でいいのにvor 2 Jahren
Idk if it tuned after I talked to it but its responds are a lot faster than 2 days ago.

anpaurevor 2 Jahren
respect for zutomayo

Benvor 2 Jahren
How much memory does this 7b model take?

La Pyae 🚀vor 2 Jahren
I wonder if it is still work for Llama 3.1 as well .

INIYSAvor 2 Jahren
To add, iPhone autocorrect feature already using smal gpt-2 based model

ShrimpApplePro 🍤 ずっと真夜中でいいのにvor 2 Jahren
Ooh interesting


