正在加载视频...

视频加载失败

HOW TO RUN AN AI MODEL FROM A USB FLASH DRIVE

94,394 次观看 • 1 个月前 •via X (Twitter)

34 条评论

Luís Rodrigues 的头像
Luís Rodrigues1 个月前

The real trick isn't the USB. It's having a model small enough to fit in memory and an inference runtime that can run locally.

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

True 😅

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

Keep in mind that while running LLMs off a USB drive works, it’s mostly meant for testing and light everyday tasks. Heavy-duty AI generation will demand a high-end system!

Vínx Corváel 的头像
Vínx Corváel1 个月前

Niiiiice

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

Try it out

Vínx Corváel 的头像
Vínx Corváel1 个月前

I legit just bought a portable SSD sandisk so I'll be setting this up hopefully this week

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

It's just for testing tho 😂

Vínx Corváel 的头像
Vínx Corváel1 个月前

nawwwww I'm doing it... and then telling everyone you told me to 🤣 I needed portable storage anyway btw lol

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

Have fun 😂😂

Rahul 的头像
Rahul1 个月前

Why do u need usb flash here?

ABDULAZIZ ALNEMASH 的头像
ABDULAZIZ ALNEMASH1 个月前

@TranslateMom arabic

The Universal Heartthrob 的头像
The Universal Heartthrob1 个月前

What's the procedure for a Linux usb or what's different in a Linux install

raymonjohns 的头像
raymonjohns1 个月前

Can it work with vscode

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

Yes it can, but it completely depends on the model size. Small quantized models (1B–3B) only need ~2–4 GB of RAM and can run in the background while coding. However, I wouldn't advise relying on local setups unless you have a high-end PC. This video is mostly for testing and running offline experiments, competent coding models need significantly more resources to be practical.

Thangchp 的头像
Thangchp1 个月前

not working!

Frank Rojas 的头像
Frank Rojas1 个月前

Thanks!

TranslateMom 的头像
TranslateMom1 个月前

With Arabic captions:

Grueber 的头像
Grueber1 个月前

@SagirGurin

Yand 的头像
Yand1 个月前

is the model free?

TheTobeGphx🖋️🇳🇬🇬🇧 的头像
TheTobeGphx🖋️🇳🇬🇬🇧1 个月前

@xdownloadercom download this please.

Moet 的头像
Moet1 个月前

Hmmm

FATÉ 的头像
FATÉ1 个月前

So far it's not unsencored it's just a turn off for me

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

???

NorthernLights.XRP 的头像
NorthernLights.XRP1 个月前

@lyonjacobsfe ...Flash Drive Modem

Assente Arbitrario 的头像
Assente Arbitrario1 个月前

@GBallarani tiè...

offNomansLand 的头像
offNomansLand1 个月前

That’s awesome! How can I do this on a Mac OS?

Madhu V 的头像
Madhu V1 个月前

Any good for working with Visual studio or vs code. As direct plugins are very slow

Ø. 的头像
Ø.1 个月前

•

nemonik 🇺🇲🐶 的头像
nemonik 🇺🇲🐶1 个月前

You can install snd keep up to data ollama and python via the scoop package manager for Windows and the huggingface cli via pip...

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

Scoop is great for managing CLI tools on a primary PC. However, it requires an internet connection to install dependencies, which defeats the point of a portable, plug-and-play USB drive designed to run 100% offline on any PC without prior setup

nemonik 🇺🇲🐶 的头像
nemonik 🇺🇲🐶1 个月前

$env:SCOOP_CACHE = "E:\ScoopCache" #drive letter of your usb drive scoop download python ... scoop install python ... mkdir E:\wheels pip download -d E:\wheels hungingface... And Bob's your uncle.

nemonik 🇺🇲🐶 的头像
nemonik 🇺🇲🐶1 个月前

That pip part is a bit fussy maybe $env:PIP_CACHE_DIR = "E:\pip-cache" pip config set global.cache-dir "E:\pip-dir "E:\pip-cache Would be better.

Axel 的头像
Axel1 个月前

How do I export the conversation to pdf? Can it?

𝐇𝐚𝐦𝐳𝐚 | Networking Guy 的头像
𝐇𝐚𝐦𝐳𝐚 | Networking Guy1 个月前

It's running on your browser so Ctrl+ P and save as pdf

相关视频