GPU.LAND offers GPU instances for Deep Learning — at dirt-cheap prices. How cheap? Up to 80% cheaper vs major cloud providers. We believe everyone, and not just "Big Tech", should be able to train AI models. So we built the cheapest place on the web to do it.
No reviews yetBe the first to leave a review for GPU.LAND
Thanks for your efforts, we are planning to use this tool extensively for CV research! Just wondering how we can use something like determined.ai for hyperparameter search by automatically spawning up many instances?
https://docs.determined.ai/lates...
@ed_muthiah Hey Ed - thanks for the kind words! As of right now, unfortunately, there is no way to spawn/shut down instances programmatically. So you would have to decide on the number of instances apriori and spin them up manually.
This can be a single instance - or if you want it can be multiple. Note that within the same gpu.land account all instances are connected inside a private VLAN. Means they can talk to each other, but no other instance on gpu.land can talk to them.
Hey gang!
@chrismessina , thanks for hunting us. 🙏
GPU.LAND is the cheapest place on the internet to train your Deep Learning models.
AI is the biggest thing that's happened to humanity, arguably, ever. Unfortunately, the current wave of AI, called Deep Learning, is VERY compute hungry. Eg the state of the art model in natural language processing, GPT3, would take about $4.6m at current GPU prices to train. Ouch.
Because model training is so expensive, it means that AI power is increasingly concentrated in the hands of "Big Tech" (Apple, Google, Facebook, Amazon), who have unlimited budgets.
GPU.LAND is a force against that. We offer some of the best GPUs for deep learning at dirt-cheap prices. We're up to 80% cheaper than major cloud providers, including AWS, GCP, Azure and others. 💸
But GPU.LAND is more than just price. We also make it super easy to start a Deep Learning-ready instance. Some highlights:
✅ Instances come pre-installed with the latest NVIDIA drivers, CUDA, cuDNN, and NVIDIA-Docker.
✅ Tensorflow, Keras, PyTorch, FastAI and other frameworks are available in separate Conda environments, ready to use.
✅ Each instance has a JupyterHub server running, serving JupyterLab over HTTPS. Launch it with 1 click of a button right from GPU.LAND's UI. Waiting time is 0.
✅ All machines in a single account are part of the same private VLAN - so you can do distributed training out of the box. Horovod is installed too.
✅ GPU.LAND sports a retro, MS-DOS-like look, because why not 😎
Pretty cool eh! Compare that to hours and hours of toil setting everything up manually at AWS/GCP... not to mention their UI is nowhere near as fun 😉
I'm around all day for AMA. 📣 Any feedback is extremely welcome. 🙏
Happy (AI) hacking!
@ilja_moi Congratulations on the launch!
GPU.LAND could be addressing the need gap - 'Democratisation of AI/ML/DL hardware' posted on my problem validation platform - https://needgap.com/problems/50-... .
You're welcomed to explain how you're solving that problem, So those who need GPU.LAND can find it easily.
Report
Been using them for the past couple weeks. Huge breath of fresh air after AWS/GCP. Instance ready in 2min instead of 2 hours of tinkering - and prices are way more affordable. Recommended for anyone in AI/ML.
Oops, their website is down? I was going to spread this good message... but when I Google it, and also find message "We cannot reach GPU.LAND right now. It is possible that they are experiencing global issues. " https://www.saashub.com/gpu-land...
I just have a doubt how they can make it so cheap... I also was told now chips are very expensive. I am looking for Kinect for while, but the whole USA does not have any. A friend from BSR told me now people sell used memory to them with a very high price too.
Report
Awesome service! Very straightforward design- you can have everything you need in a single click, or 3 if you need the full suite. I was able to generate this image in 15-20 minutes using one v100!
Report
Simple, practical and doesn't require certificate-level training to use it to the fullest. I already recommended it to my colleagues.
Report
Love how simple it is it setup and use, it is a game changer for quick and easy experimenting, I was able to spin up a few instances, use Jupyter lab straight away, clone my repo, and begin training an intensive ML model in a few minutes. Trained a VAE and a GAN
GPU.LAND
GPU.LAND
Bufferi.ng
Bone Health Tracker
GPU.LAND