Liam Schoneveld
Liam Schoneveld
Just had the same issue. I think the warning occurs when you call `opt.swap_swa_sgd()` before any stochastic weight averaging has actually taken place. Here's my toy example: ```python import torch...
I made a fork here which should make things clearer on how to run this on mac https://github.com/nlml/vadnet
@khakhulin Is there a way to reduce this jitter? What is the 'camera from the mobile-net' that you refer to?
> Only thing I'm not sure about is whether, if you have multiple outputs, the outputs in `current_output_names` are ordered according to the order as defined in the `forward` function...
Hmmm.. seems like the fnames list might be empty? Perhaps look into how that is created and see why it might be empty...
FYI: I ran into the same problem on a machine with a 3090 (24gb GPU mem) and 64GB of RAM. Changed `load_full_step: 8` to `load_full_step: 16` and `subsample_frac: 0.125` to...
@ookey thanks a lot! That is indeed very helpful! Will report back if this leads to improvements.
@breuckelen Thanks for getting back to me! Basically the best results that I've gotten so far are just running `run_one_n3d.sh`, which has NDC. The only changes made are the image...
Thanks again for your all your comments/advice. I tried the `white_bg=1` but it actually made things worse. I've managed to improve my best validation PSNR from about 28 to about...
@seva100 Would you be able to provide a short example script of how to load one of these checkpoints found in that Google Drive? I tried quite a few things...