Small change in image swap scripts. #267
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Hey everyone, i dont know how if everyone here already fixed the image swap files so that you dont get a bunch of warnings when you use it. Also renaming the image results to anything else doesn't work. Ive made the necessary corrections and put in a pull request. If anyone here who's not too familiar with python, would like me to post the changes, let me know and ill post it to my repo. Cant wait to see what the simswap team is working on!!
I don't know much about python, and I'm about to give up because of the errors It would be greatly appreciated if you could post your modifications
Sure bud, will post it tonight or in the morning. What are you giving up on? What problem are you having?
thanks for your contribution, I trained the new 512 model, but it doesn't reason correctly, if you're in the mood I can share my checkpoint
No problem brother, contributing is the best part of github. On my github page, today i will be posting some alterations and solutions to various simswap related issues. One, that for most users will make the script a bit faster (hopefully) and the image swapping without the warning and being able to name the image whatever you want and save it to a folder of your choice. I can certainly take a look at your checkpoint. I haven't gotten around to training a model yet because ive been focused on optimization but im sure we can figure out what the issue is. My email is darumperez@gmail.com if you need to send anything directly. And if you haven't already done so, I suggest joining discord channel that's been created. Good luck, and no reason to give up; we're all here to help eachother out.
I am using an RTX A5000 with 24GB RAM but, inference is still slow at 3 to 4 it/s. It starts with 24 it/s than climbs down to 3 to 4 it/s after few seconds. GPU utilization is also low. Do not know how to fix this.
Hey, at first glance that seems to be a bottle neck issue. It would help if you can list your environment and confirm that you have the proper versions of pytorch and onnxruntime-gpu installed. The version needs to be compatible with your native version of Cuda.
Bro, did you post your update? I'm excited to see it
Can you open an issue to explain your problem, please ?
Posting today!! Once I get home :)
Hi, have you evaluated the generated results? I evaluated with https://github.com/MuggleWang/CosFace_pytorch, but got bad id_retrieval