deepfates
deepfates
This looks quite useful, thank you! It can do CRUD ops on individual elements in the index which is desirable. It does look like it needs to know the maximum...
Yes! Ideally, I'd like to be able to run a server that can track databases by user. So there could be Android/iPhone clients etc. without having to run giant PyTorch...
Wow, I've never seen this before. Have you watched your RAM increase during this process? Does it top out your memory? Definitely need better support for resuming index. I am...
I'm afraid that it's probably batch 264 rather than a specific image. The data loader is a pie torch feature on specifically using for batch processing, so it's doing your...
That could work! Still doesn't explain why that one batch uses so much memory but could get you through the bottleneck for now. I think I hard-coded some batch and...
Yeah, i ran across this too. Thanks for recording it On Tue, Mar 22, 2022, 10:03 PM wkrettek ***@***.***> wrote: > Memery serve will break if you don't give a...
Thanks! I will test this out. I think there was a problem with using external repos from settings.ini but I don't remember why. Certainly don't want to drag in anything...
I see that the official CLIP repo recommends using Torch 1.7.1. Have you been using it fine with a different version @russelldc ?
Okay, I have updated the dependencies. Found that PyPi doesn't allow `install_requires` to have dependencies on external repos, like "clip@git+https://github.com/openai/CLIP.git", but instead can use `dependency_links` which is built from `dep_links`...
Scratch that, it seems this doesn't work due to many reasons including the fact there's already something called `clip` in PyPi. Tried converting it to huggingface repo but they use...