@forcceee_: Soup just broke fine-tuning barriers by enabling large language model training on low VRAM devices using as-layer streaming. This open-source innovation makes AI democratization real. #souptrain #languagemodels #finetuning #aiinnovation #opensource

Derek Holloway

Open In TikTok:

Region: PT

Saturday 12 September 2026 19:40:01 GMT

20672

1035

370

Music

Download

No Watermark .mp4 (8.61MB) Watermark .mp4 (8.99MB) Music .mp3

Comments

1 :

Anybody try running Claude using this ?

2026-09-24 01:22:32

YasinsGlobe :

I tried it on an RTX 4500 Qwen2.5-7B LoRA pretty good results

2026-09-13 08:13:23

King Ragz :

This video is a ai workflow everything is these days. But sweeet. I’m gonna check it out

2026-09-14 15:26:33

Wak Anda For Eva :

borat [Heart eyes]

2026-09-23 15:21:51

skip :

Does Nvidia know about this? Hope he stays safe and alive 😂😂😂

2026-09-13 00:51:09

LJ :

You don’t need to fine tune models anymore

2026-09-15 14:25:47

BL :

This is good for fine-tuning, but not for running the model… You still need VRAM…

2026-09-16 05:38:06

CapcsellCoreConsciousnessArmor :

thats one of my code systems they stole

2026-09-14 19:04:50

RizbIT :

how do I use it to run a model locally. which models does it work with

2026-09-13 10:45:22

Dark_Sense :

So, it's basically what Fizgig is for Krea2 T2I & H3... 🤔

2026-09-13 09:36:14

Francesco Spycher :

awesome

2026-09-14 03:46:58

Twisted Grave :

I would like a link to the GitHub plz

2026-09-13 10:34:49

BlerdHulk :

Trying to tank NVIDIA stock

2026-09-16 14:23:02

To see more videos from user @forcceee_, please go to the Tikwm homepage.