Speed and model size

[initialneil]

Hi, thanks for the sharing. I tried to run the infer.py on A100 and it takes 48 seconds with 36 steps.
While the project page claims 1 second inference time and 960M backbone. Is the model not released yet?
I see the file on hf is 4GB: https://huggingface.co/WeiChow/EditMGT/tree/main/transformer

[mmlingyu]

also the problem , with rtx 4090 and 36 steps , it cost 12s.

[Peters213]

the same problem , with A100 and it takes 12 seconds with 36 steps