MCPcopy Create free account
hub / github.com/Zhiyuan-R/Tiger-Diffusion / forward

Method forward

modules/pointnet.py:32–43  ·  view source on GitHub ↗
(self, inputs)

Source from the content-addressed store, hash-verified

30 self.mlps = nn.ModuleList(mlps)
31
32 def forward(self, inputs):
33 features, coords = inputs
34 if self.include_coordinates:
35 features = torch.cat([features, coords], dim=1)
36 coords = torch.zeros((coords.size(0), 3, 1), device=coords.device)
37 if len(self.mlps) > 1:
38 features_list = []
39 for mlp in self.mlps:
40 features_list.append(mlp(features).max(dim=-1, keepdim=True).values)
41 return torch.cat(features_list, dim=1), coords
42 else:
43 return self.mlps[0](features).max(dim=-1, keepdim=True).values, coords
44
45 def extra_repr(self):
46 return f'out_channels={self.out_channels}, include_coordinates={self.include_coordinates}'

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected