search

OpenGVLab / unmasked_teacher

[ICCV2023 Oral] Unmasked Teacher: Towards Training-Efficient Video Foundation Models
https://arxiv.org/abs/2303.16058
MIT License
285 stars 15 forks source link

Inference cost #11

Closed tengyu-liu closed 1 year ago

tengyu-liu commented 1 year ago

Great work! The performance on public sources is very impressive. Congratulations!

The paper did not mention the cost of inference. How many A100s do you need for inference?

Andy1621 commented 1 year ago

It only requires one A100 for inference since the large model is only ViT-large. For retrieval, if your dataset is too large, maybe you need more GPUs to have a quick inference.

tengyu-liu commented 1 year ago

Thank you very much. I'm closing this issue.