Sakr
Sakr
> > val_err_tot > > This seem to solve that problem miss match between two mel spec, but when I compare difference y_mel, y_g_hat_mel, It is larger than when using...
> Updated weights (without typo) and mixed precision support now at [this fork](https://github.com/L-Reichardt/Cylinder3D_spconv_v2) I got this error while using your code 
> @Body123 The error is caused by the versions of CUDA / SpConv. The code should run regardless. No I check all version required and the error still happen ,...
but what is the difference between these: 2024-04-23 15:23:41.869548: Pseudo dice [0.8399] 2024-04-23 15:23:41.911069: Yayy! New best EMA pseudo Dice: 0.8196 each one of them is related to training and...
Okay thanks alot, my last question, and sorry for many questions: those numbers are based on taking some random patch from the training set during epoch not the whole image...
did you reach anything?
because I want to predict a huge a mount of images and the inference is too slow, can you suggest something that can help me, please?
is there any updates
or is there any similar command that helps us with more batch size rather than one?
increase the -num_parts and -part_id needs multiple GPUs?