garvita-tiwari/PoseNDF

posendf's forward function self.model_occ

2019EPWL opened this issue · 4 comments

hi, in the posendf.py there is a dist_pred = self.model_occ(rand_pose.reshape(self.batch_size,84)), but I can't find the definition of self.model_occ in the repo, could you help with it?

rczDS commented

i've got the similar quetion, in posendf.py:
self.dfnet = DFNet(opt['model']['DFNet']).to(self.device))
dist_pred = self.dfnet(rand_pose_in.reshape(self.batch_size,84))
but in default amass.yaml, "DFNet" is not defined under "model". anybody knows the params?

Besides, the pretrained model is very strange, 1. why it has conv layers? 2. it has no dfnet parameters 3. it has no StructureEncoder parameters. (I change the model.CanSDF to model.DFNet and model.StrEnc.use to True)
1
2

same problem

please use the latest commit with updated config file: 4b8f1f0