Furkan Gözükara
patreon
How To Deploy A Pod On RunPod And Verify It Is Working
🕑 Added 2024-02-05 21:08:39 +0000 UTCJoin discord and tell me your discord username to get a special rank : SECourses Discord
In this post I am going to show you how to select your template and setup your Pod accurately with the newest interface. Moreover, I am going to show you how to ensure the rented pod is working accurately.
First register your RunPod account if you haven't yet.
You can use this link (help me to get credits so I can do more research and tutorials) : https://runpod.io?ref=1aka98lq
Then use this link to search for any template you want to use : https://www.runpod.io/console/explore
After selected your template you will get to next page where you will see template explanation and Deploy GPU button
Next you can select Secure Cloud or Community Cloud. I prefer to use Community Cloud for majority of tasks
After you selected your GPU, you can Customize Deployment and also select On-Demand or Spot (bidding based) renting.
You can add new port to Customize Deployment to connect Gradio apps. By default Gradio apps starts at 7860 port so you can add 7861 port to connect that started Gradio app through proxy of RunPod.
You can also increase your Volume Disk which is persistent and will remain until you delete your Pod. Your Container disk space is temporary and whatever you put there will be reset after you stop your pod.
After everything is set, click continue, and then verify everything in last screen and click deploy.
After click deploy, go to your Pods link : https://www.runpod.io/console/pods
And there you will see your Pods.
Then wait until connect button appear click it and connect to Jupyter lab interface.
After you Jupyter Lab interface loaded, upload attached verify.ipynb into workspace, double click it and click play icon. Check the printed message. If you are not getting a message like shown below, Pod is broken. So delete it and move to another Pod.
The code of ipynb notebook is as below
import torch
# Check if CUDA (GPU) is available
if torch.cuda.is_available():
# Get the number of available CUDA devices
num_cuda_devices = torch.cuda.device_count()
print(f"Number of CUDA devices available: {num_cuda_devices}")
# Print information about each CUDA device
for i in range(num_cuda_devices):
device = torch.device(f'cuda:{i}')
print(f"CUDA Device {i}: {torch.cuda.get_device_name(i)}")
print(f"Device properties: {torch.cuda.get_device_properties(i)}")
# Set a default device to the first CUDA device (GPU 0)
torch.cuda.set_device(0)
# Verify if CUDA is working by performing a simple operation on the GPU
a = torch.tensor([1.0], device="cuda")
b = torch.tensor([2.0], device="cuda")
c = a + b
print(f"Result of GPU computation: {c.item()}")
else:
print("No CUDA devices available. Make sure you have a compatible GPU and PyTorch with CUDA support installed.")
Comments
Furkan Gözükara
I never used that template. I use either pytorch, stable diffusion templates. I have auto installer for Kohya SS GUI (installs latest version) : https://www.patreon.com/posts/for-runpod-kohya-84898806
Thomas
I just tried it with the KohyaSS runpod template and got "ModuleNotFoundError: No module named 'torch'"
Furkan Gözükara
yes you can run all. start with stable diffusion template use our auto installers they make unique venv https://www.patreon.com/posts/fooocus-sd-web-1-92759045 https://www.patreon.com/posts/runpod-comfyui-86062569
Pallavi Chauhan
thanks for making this detailed guide, I could never find the bidding option in this new interface. Is there a way to run comfyui, A1111, and foocus all at once in one pod?
