Jose Javier
Jose Javier
Would you be open to having a page on Condition Variables, i.e. `sync.Cond`? It could go right after the mutex page and I believe it'd be a great resource since...
There is a substantial loudness difference between youtube videos. This is specially egregious when comparing music videos and more general types of content. Music videos are consistently louder, sometimes to...
I understand that you can not share the fonts due to copyright issues. Could you comment which fonts did you use for the training? Thank you
Hi, thanks for creating komorebi, it's the best tiling window manager I've come across. I've been a yabai user for a while and there are a few workflows I'm sorely...
I'm in working with a distributed training setup and need to make it work with cluster preemption. I've read the Checkpointing docs but those do not address what happens when...
Expanding out the terms, the NCC loss computational graph can be simplified. Tested with random inputs and got equivalent results with both implementations
PyTorch added the functionality to have non-persistent buffers that are not included in the state_dict [1]. This is done by specifying `persistent=False` when registering it. This is just an improvement,...
I'm unable to `pip install` the package. My current environment is Ubuntu 18.04.4 with versions ``` cudatoolkit 10.1.243 h6bb024c_0 pytorch 1.6.0 py3.7_cuda10.1.243_cudnn7.6.3_0 pytorch torchvision 0.7.0 py37_cu101 pytorch ``` The error...
First, thanks for developing this great library I was digging through the composer docs because I wanted to test out blurpool. While the [Method entry](https://docs.mosaicml.com/en/stable/method_cards/blurpool.html) is great (has examples for...
Trying to decide how to do 5-fold cross-validation with FFCV loading. Does FFCV have any way of supporting loading just a subset of the indices, or should I have 5...