Software Engineering Courses (SECourses)•10mo ago

1350Megabit down, 35Megabit up

computersheepOP•3/5/25, 8:22 PM

and yes the download does go slightly above 1Gb...., I likely have the only house on this internet provider that even supports using the extra 100Mb download lol

Ccomputersheep 1350Megabit down, 35Megabit up

Furkan Gözükara SECourses•3/5/25, 8:23 PM

wow upload worse than mine

Furkan Gözükara SECourses•3/5/25, 8:23 PM

so i guess same where you live

computersheepOP•3/5/25, 8:23 PM

just my neighboord, 4 different fiber providers in this city, many offering symmetric 5Gb up/down if your willing to pay for it

computersheepOP•3/5/25, 8:24 PM

Verizon offered to run me a dedicated line for just 14k/month

FFurkan Gözükara SECourses i wonder why thinlinc client cant make any proper data transfer

computersheepOP•3/5/25, 8:26 PM

I think it relates to ping/latency. which is high for people in Europe trying to get to our datacenters.

Ccomputersheep Verizon offered to run me a dedicated line for just 14k/month 😦

Furkan Gözükara SECourses•3/5/25, 8:27 PM

14k month lol

Ccomputersheep I think it relates to ping/latency. which is high for people in Europe trying to...

Furkan Gözükara SECourses•3/5/25, 8:28 PM

i believe in our case it is purely to sell you high upload from like 100x price :d

computersheepOP•3/5/25, 8:28 PM

Yea it was for a dedicated 1Gb line

computersheepOP•3/5/25, 8:28 PM

but 1Gb is nothing for that price...

Furkan Gözükara SECourses•3/5/25, 8:29 PM

that price is just insane

computersheepOP•3/5/25, 8:29 PM

Yea I coudln't afford it even if I was crazy enough to consider paying that amount

Furkan Gözükara SECourses•3/5/25, 8:30 PM

yep. in Türkiye they sell such upload from huge prices to businesses. sole reason to keep upload of consumers lower. so you have to pay more

Furkan Gözükara SECourses•3/5/25, 8:30 PM

like universities paying huge money

Furkan Gözükara SECourses•3/5/25, 8:31 PM

like nvidia abusing monopoly

Zaulinator•3/5/25, 8:54 PM

yeah its not to bad here in the US for fiber. 70 dollars for 1gb up/down

computersheepOP•3/5/25, 9:14 PM

Anything I need to do to make sd_wan use multiple GPUs?

Ccomputersheep Anything I need to do to make sd_wan use multiple GPUs?

Furkan Gözükara SECourses•3/5/25, 9:22 PM

sadly not possible yet

Furkan Gözükara SECourses•3/5/25, 9:22 PM

i requested this but they said not priority

ZZaulinator yeah its not to bad here in the US for fiber. 70 dollars for 1gb up/down

Furkan Gözükara SECourses•3/5/25, 9:22 PM

great

Furkan Gözükara SECourses•3/5/25, 9:23 PM

i had to do forced shut down pc literally became frozen due to xformers compile i presume

Furkan Gözükara SECourses•3/5/25, 9:23 PM

i will leave it during sleep

computersheepOP•3/5/25, 9:27 PM

computersheepOP•3/5/25, 9:28 PM

I guess ill have to live with 35.52s/it on one gpu

computersheepOP•3/5/25, 9:37 PM

hmmm the github implies multi gpu is supported
https://github.com/Wan-Video/Wan2.1

Multi-GPU inference using FSDP + xDiT USP

pip install "xfuser>=0.4.1"
torchrun --nproc_per_node=8 generate.py --task t2v-14B --size 1280*720 --ckpt_dir ./Wan2.1-T2V-14B --dit_fsdp --t5_fsdp --ulysses_size 8 --prompt "example"

pip install "xfuser>=0.4.1"
torchrun --nproc_per_node=8 generate.py --task t2v-14B --size 1280*720 --ckpt_dir ./Wan2.1-T2V-14B --dit_fsdp --t5_fsdp --ulysses_size 8 --prompt "example"

Ccomputersheep hmmm the github implies multi gpu is supported https://github.com/Wan-Video/Wan2...

Furkan Gözükara SECourses•3/5/25, 9:38 PM

yes but i dont use it atm

Ccomputersheep hmmm the github implies multi gpu is supported https://github.com/Wan-Video/Wan2...

Furkan Gözükara SECourses•3/5/25, 9:40 PM

Furkan Gözükara SECourses•3/5/25, 9:40 PM

leave a comment

https://github.com/modelscope/DiffSynth-Studio/issues/376

GitHub

Can we utilize multiple GPU to speed up single video generation wit...

The pipeline working great excellent work you got there Can we utilize multiple GPU to speed up? any example? Thank you so much

Furkan Gözükara SECourses•3/5/25, 9:41 PM

Furkan Gözükara SECourses•3/5/25, 9:44 PM

wan 2.1 training coming

Furkan Gözükara SECourses•3/5/25, 9:59 PM

i am gonna add save load config now before sleep

Furkan Gözükara SECourses•3/5/25, 10:13 PM

RTX 4090 FLUX DreamBooth / Finetuning speed - 0 VRAM usage having GPU

FFurkan Gözükara SECourses RTX 4090 FLUX DreamBooth / Finetuning speed - 0 VRAM usage having GPU

Timson•3/5/25, 10:20 PM

wait, what? this is faster than l40s

Furkan Gözükara SECourses•3/5/25, 10:23 PM

TTimson wait, what? this is faster than l40s

Furkan Gözükara SECourses•3/5/25, 10:23 PM

it shouldnt be than L40S

xlads•3/5/25, 10:30 PM

in "Solving CUDA and C++ Tools Compiling Errors Once and For All for Python AI Applications" i notice there is 3 CUDA versions is it necessary to download them all?

xlads•3/5/25, 10:31 PM

12.8, 12.4 & 11.8

Furkan Gözükara SECourses•3/5/25, 10:35 PM

https://www.reddit.com/r/SECourses/comments/1j4g67d/new_amazing_updates_arrived_to_our_ultimate_wan/

From the SECourses community on Reddit: New amazing updates arrived...

Explore this post and more from the SECourses community

Xxlads in "Solving CUDA and C++ Tools Compiling Errors Once and For All for Python AI A...

Furkan Gözükara SECourses•3/5/25, 10:35 PM

cuda 12.8 should be sufficient atm

Furkan Gözükara SECourses•3/5/25, 10:46 PM

i slept like 3 hours yesterday i have to sleep asap

xlads•3/6/25, 12:04 AM

Does anyone know if live portrait is still best for lip-syncing, also can i train a flux with just one image? im trying to create an ai influencer but not sure how to genrate multiple pictures other then doing faceswaps

�𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮 There are a few people on reit in the LocalLama subreddit who have bought these ...

VAMP•3/6/25, 2:03 AM

can you share the subreddit?

blackhat•3/6/25, 4:22 AM

@Dr. Furkan Gözükara you are using this in your prompts a lot what is this assosiated to , is it for some particular expression of yours or the some part of your face it is mentioned for ?

xlads•3/6/25, 7:01 AM

got this error while trying to install Batch Image Pre-Processing Gradio APP for Windows "ERROR: THESE PACKAGES DO NOT MATCH THE HASHES FROM THE REQUIREMENTS FILE. If you have updated the package versions, please update the hashes. Otherwise, examine the package contents carefully; someone may have tampered with them.
gradio from https://files.pythonhosted.org/packages/dd/45/64acd7bed4abe64358c8ffa5256d6d25f54cc7b87c43fcaeda1b94ba243e/gradio-5.20.0-py3-none-any.whl (from -r requirements.txt (line 10)):
Expected sha256 85e649611c343fa6c2cff50e3f6a662046158bf592eb954dffdca24e896185e5
Got 0731aec925fbd2dfad0808d351fc16d2c261932346f2944e4636eca877f402c6"