dbf6147504
The logic is that the discriminator may be incapable of providing a truly targeted loss for all image regions since it has to be too generic (basically the same argument for the switched generator). So add some switches in! See how it works! |
||
---|---|---|
.. | ||
__init__.py | ||
convert_model.py | ||
distill_torchscript.py | ||
gpu_mem_track.py | ||
numeric_stability.py | ||
onnx_inference.py | ||
util.py |