训练迭代14500次,为什么效果很差还存在多面问题?
jim-1ee opened this issue · comments
您好,我这边也想跑跑这个代码,但是我出了点问题,想咨询一下可以么,有偿。主要是这个BUG:
DeprecationWarning: Type google.protobuf.pyext._message.ScalarMapContainer uses PyType_Spec with a metaclass that has custom tp_new. This is deprecated and will no longer be allowed in Python 3.14.
您好,我这边也想跑跑这个代码,但是我出了点问题,想咨询一下可以么,有偿。主要是这个BUG: DeprecationWarning: Type google.protobuf.pyext._message.ScalarMapContainer uses PyType_Spec with a metaclass that has custom tp_new. This is deprecated and will no longer be allowed in Python 3.14.
Just a warning, should be ok for your training process
Same here, 15000 iterations still occur severe Janus problem!
Same here. tried several prompts, such as corgi, panda, and man in a suit. All occurred severe Janus problem
--- update ---
the previous poor results were because when executing this command:
python main.py --config-name=base prompt.prompt="<prompt>"
the prompt for point-E was set to "a human face" as defined in base.yaml
.
Instead, after specifying init.prompt="<point-e prompt>"
, I got good results without Janus problem.
BTW, it seems that the default configuration (conf/base.yaml
) does not enable full functions, particularly the compactness-based densification and pruning, which are major contributions of the paper. Why are they not enabled? Wouldn't this lead to sub-optimal results?