caphohotain

joined 10 months ago
sorted by: new top controversial old
Anyone running 3 gpus? Looking for advice on best x670 that might be able to slot a third card on. in c/localllama@poweruser.forum
[–] caphohotain@alien.top 1 points 9 months ago

A graphics card of any kind won't fit in the bottom PCIe slot because of all the io connectors next to it. You need a riser yo use the bottom PCIe slot.

Serious inquiry: I've been tinkering a lot with finetuning and was wondering if it would be worth to buy a V100 of my own in c/localllama@poweruser.forum
[–] caphohotain@alien.top 1 points 9 months ago

This. I was so confused when I saw op's post: why on earth to buy an old only 16gb vram card with the price of multiple larger vram and newer cards?

Anyway to save your cloud GPU fine-tuned models to your local storage? in c/localllama@poweruser.forum
[–] caphohotain@alien.top 1 points 10 months ago

Thanks! I will search for SFTP and S3:)

1
Anyway to save your cloud GPU fine-tuned models to your local storage? (alien.top)
 

I want to fine tune some open source LLMs on cloud GPUs (Google, runpod etc.), but most of the tutorials I read is to upload the fine tuned models to Huggingface hub. I want to keep them privat. What can I do? Thanks!

Clearing up confusion: GPT 3.5-Turbo may not be 20b after all in c/localllama@poweruser.forum
[–] caphohotain@alien.top 1 points 10 months ago

It's 20b or .2b or 200000000b doesn't bother me at all.