Victor
Victor
Yep, but it turns out ffmpeg is very inefficient because gdigrab is an old gen tech. These days the desktop is in DWM DXGI buffers, and there's an efficient API...
Just faced the same problem. Generally, .NET apps must store secrets in `System.SecureString` or their own implementations of a similar concept. Secrets should be stored in an unmanaged memory (because...
@AArnott just an opinion, call it an educated guess if you will. TL;DR; I am concerned about who will maintain Windows.Win32.winmd and CsWin32. I wouldn't want this project to switch...
@fbcotter any plans to release 1.3 with 1D DWT?
@lostmsu no, I mean PyPi only has 1.2, so to get 1.3 one currently needs to install from source.
Just curious, with the current version, can I even do minibatches?
All right, from Classify2DDemo I guess I just need to set the last parameter of Volume's Shape to number of items in the minibatch. Can you confirm that?
@cbovar IMHO, you should mention that in README. Seems like an important detail.
Sounds like it would make sense to distribute larger models in binary format instead of PyTorch format. Right now offload scenarios double the storage requirements due to that issue.
Alternatively, could you make it possible to download the model in PyTorch format, then convert it to binary for `accelerate` to read directly from "offload_folder"?