DEKUBE NETWORK(@dekube_official) 's Twitter Profile Photo

Sometimes we are moved by the Open Source spirit of people like Alignment Lab AI , and we will aggregate larger compute power to support these very good Hugging Face community project!

Sometimes we are moved by the Open Source spirit of people like @alignment_lab , and we will aggregate larger compute power to support these very good @huggingface community project! 

#gpupoor #OpenSource #llama #compute
account_circle
Vanya 2V(@vanya2v) 's Twitter Profile Photo

Our new MV 'Bismillah' is OUT NOW!

I include several AI animation techniques: Video Inpainting, vid2vid to anime & papercraft style, and Deforum SD, also standard im2vid.

I had fun combining real footage with AI video and , though I'm

account_circle
Saravana Rathinam(@skr_elbo) 's Twitter Profile Photo

The Puppetry (closet) Datacenter 🦩running hot yesterday at 107 degrees.

All GPUs going 💨, serving a user a minute 📈

The Puppetry (closet) Datacenter 🦩running hot yesterday at 107 degrees.

All GPUs going 💨, serving a user a minute 📈

#gpupoor
account_circle
Xavier Manuel Mountain(@XaevrM) 's Twitter Profile Photo

This tells you enough when I say the term and

Tech industry is on a whole different level of what rich means.

If you have a few millions as a startup that is a cute number.

It takes massive funding to be able to survive the echo system of big tech.

Why

This tells you enough when I say the term #gpupoor and #gpurich

Tech industry is on a whole different level of what rich means.

If you have a few millions as a startup that is a cute number.

It takes massive funding to be able to survive the echo system of big tech.

Why
account_circle
Ilyas Abdourahim(@AbdourahimIlyas) 's Twitter Profile Photo

Asking Copilot a Fermi-estimation of Flops and Devs available for OpenSource, and ChatGPT3.5 requirement.

OpenSource capacity
FLOPs per month = 1e+23 FLOPS
Devs-time per month = 1e+9 hours

ChatGPT3.5 on 4month
FLOPs = 1e+24 FLOPS
Devs-time = 1e+5 hours

OpenSource is

Asking Copilot a Fermi-estimation of Flops and Devs available for OpenSource, and ChatGPT3.5 requirement.

OpenSource capacity
FLOPs per month = 1e+23 FLOPS
Devs-time per month = 1e+9 hours

ChatGPT3.5 on 4month
FLOPs = 1e+24 FLOPS
Devs-time = 1e+5 hours

OpenSource is #GPUpoor
account_circle
notcollinalexbell(@collinalexbell) 's Twitter Profile Photo

redlining my Intel integrated GPU to generate this blurry stable diffusion cat with tinygrad. Its nice to having OpenCL as a first class citizen!

redlining my Intel integrated GPU to generate this blurry stable diffusion cat with tinygrad. Its nice to having  OpenCL as a first class citizen!

#gpupoor
account_circle
Ä̷̬͖̽͗P̷̭̳͔͇̊ on CompSciFutures (∀/acc)(@CompSciFutures) 's Twitter Profile Photo

Other alternatives include:

- The Global Optimisation family of algorithms (e.g., Simulated Annealing)

- Putting a rejection sampler over your learning rate, so you have MCMC Gradient Descent learning



.\p

Other alternatives include:

- The Global Optimisation family of algorithms (e.g., Simulated Annealing)

- Putting a rejection sampler over your learning rate, so you have MCMC Gradient Descent learning

#BasementAGI #GPUPoor #NeuralNetworks 

.\p
account_circle
mateusz dziemian(@mattmdjaga) 's Twitter Profile Photo

Got my Hugging Face merch for helping out in the upcoming community computer vision HF course. Thanks merve and @johko990 for organising the course 🤗.

Got my @huggingface merch for helping out in the upcoming community computer vision HF course. Thanks @mervenoyann and @johko990 for organising the course 🤗.
#huggingface #gpupoor
account_circle
Brice Gower(@BFresh256) 's Twitter Profile Photo

Augment has always built and maintained our own compute for developing Machine Learning algorithms; currently utilising 3x RTX3090's, 2x 2080ti's, and 5x 1070's

Being forced us to focus on algorithmic improvement, leading to our best innovations.
bit.ly/3OUCAdP

Augment has always built and maintained our own compute for developing Machine Learning algorithms; currently utilising 3x RTX3090's, 2x 2080ti's, and 5x 1070's

Being #GPUpoor forced us to focus on algorithmic improvement, leading to our best innovations.
bit.ly/3OUCAdP
account_circle
Vicky S(@VickyLovesAI) 's Twitter Profile Photo

My

There is a new AI chip by called Inferentia

It brings in significant cost saving when compared to Nvidia gpus



aws.amazon.com/blogs/aws/amaz…

My #gpupoor #indiehackers 

There is a new AI chip by #AWS called Inferentia

It brings in significant cost saving when compared to Nvidia gpus

#buildinpublic 

aws.amazon.com/blogs/aws/amaz…
account_circle