PointNetSetAbstraction question

[Lemon2048]

Hello, I made a small modification in the class PointNetSetAbstraction(nn.Module) by adding 1 to the variable in_channel. After making this change, I encountered an issue in the subsequent code:

python
for i, conv in enumerate(self.mlp_convs):
bn = self.mlp_bns[i]
new_points = F.relu(bn(conv(new_points)))

After executing this code, all the values in new_points turn into NaN. I would like to inquire about what might have gone wrong. Thank you.

[abysee]

It is likly that there's a line before calling PointNetSetAbstraction that it sets defalt input channel to 6 or 3. Like in_channel = 6 if normal_channel else 3. You may have to modify the corresponding line.