After the possibility of May on the run -pod platform, I started searching for other suppliers. Vast Offers RTX4090 for the under Hour 0.49 per hour This means that it is much cheaper than using 20 cents on a run pod. So as my first experience, I decided to see if I could be trained by Flox Laura using AI-Toolkit UI. I have found which templates will be best suitable and in this post there are steps to work extensively to AI-Toolkit UI.
Create an account first Vast And load some credit, I started with $ 20 for all my experiences.
Select the template
The template you want to use is their PyTorch (wide)Set the disk space to allocate about 80 GB, about 80 GB. You will see the GPUS available on the right and you can filter to select RTX 4090. You see in the screenshot that they start at least $ 0.335/hour, you also see where they will go. Hong Kong, Spain, Utah etc.
For example, once you start, you can access your example from the example menu. You can take further steps for example through the toolbar at the bottom of the example. Don’t worry if you look different, the reason for this is that I started using the latest UI (Settings> Early Access> Opt -InJes
Click to open applications Open And then you can launch Jupyter request. Something else you should see which one is Tunnels. We will use it later, thus we expose additional applications through their specific ports.
Now that you have Jupyterlab running Open the Jupiterlib There are all the bells and whistles that we need to install AI-toolkit.
We are now ready to start installing the AI-Toolkit app, which takes only a few minutes.
Install AI-toolkit on the wide
Open Terminal I Jupyterlab And just copy the paste to the following command to run. The beauty of Vast The template is that it has already been installed NPM (node.js) so we do not have to do extra work.
git clone
cd ai-toolkit
git submodule update --init --recursive
python3 -m venv venv source venv/bin/activate
pip3 install --no-cache-dir torch==2.6.0 torchvision==0.21.0 --index-url
pip3 install -r requirements.txt
Once it is installed and complete, you need to set up a new tunnel and then launch the app.
Remember the top to configure the tunnel Tunnel Section AI-TOKIT runs at the UI port 8675. Enter the tunnel address And click Create a new tunnel. Setup takes a few seconds, so make sure you do this before launching AI-toolkit UI.
Open a new Terminal Leave just the following command to launch the window and AI-toolkit UI in Jupyterlab
cd /workspace/ai-toolkit/ui
npm run build_and_start
Once the service and application is ready and running, you can copy the URL of the tunnel and open it to the new browser tab/window. You have an osteress – ready to go to the AI ​​toll cut.
Training your Laura L you need to make and upload yourself Datasis. Next Setup A New job The settings you want and you can go to sleep. If you are going to train on the flow, make sure you update the hugging face to the token Settings Page
If you want to support our site please consider buying us FiFor, for, for,. Catch a product Or Subscribe. Fast GPU is required, access the fastest GPUS for less than $ 1 per hour runpod.io