Jupyter Keeps crashing when trying to initialize CLIP

[SSdevelop]

I have the following code in a jupyter cell:

device = "cuda" if torch.cuda.is_available() else "cpu"
model, preprocess = clip.load("ViT-L/14", device=device)

My kernel keeps crashing. I have tried to run the same cell on Google Colab as well (Using T4 GPU with high RAM). Has anyone faced the same issue? Message from local Jupyter Lab: Message from Google Colab: App Logs from Google Colab:

[SSdevelop]

Turns out if I remove decord, it works perfectly fine. Anyone knows why?

[rakhon9]

Not really a solution, but I have come accross this issue before. My kernel kept dying. It turned out it was because of conda environment. I accidentally installed some library versions that crash with each other. I have solved the problem by creating a brand new conda environment and only installing required libraries for CLIP running. I hope it helps.