Hi Simar!Your work is very great!!
When I tried to reproduce the code of "VinL",I can finish the locomotion part,including training and test.
But when I tried to use the navigation policy trained by myself,following the instruction in https://github.com/joannetruong/habitat-lab/blob/kin2dyn/README.md,it does not work.The report is followed:
RuntimeError: Error(s) in loading state_dict for PointNavResNetPolicy: size mismatch for net.prev_action_embedding.weight: copying a param with shape torch.Size([32, 3]) from checkpoint, the shape in current model is torch.Size([32, 2]). size mismatch for net.tgt_embeding.weight: copying a param with shape torch.Size([32, 4]) from checkpoint, the shape in current model is torch.Size([32, 3]). size mismatch for net.visual_encoder.compression.0.weight: copying a param with shape torch.Size([102, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([137, 256, 3, 3]). size mismatch for net.visual_encoder.compression.1.weight: copying a param with shape torch.Size([102]) from checkpoint, the shape in current model is torch.Size([137]). size mismatch for net.visual_encoder.compression.1.bias: copying a param with shape torch.Size([102]) from checkpoint, the shape in current model is torch.Size([137]). size mismatch for net.visual_fc.1.weight: copying a param with shape torch.Size([512, 2040]) from checkpoint, the shape in current model is torch.Size([512, 2055]). size mismatch for action_distribution.mu.weight: copying a param with shape torch.Size([3, 512]) from checkpoint, the shape in current model is torch.Size([2, 512]). size mismatch for action_distribution.mu.bias: copying a param with shape torch.Size([3]) from checkpoint, the shape in current model is torch.Size([2]). size mismatch for action_distribution.std.weight: copying a param with shape torch.Size([3, 512]) from checkpoint, the shape in current model is torch.Size([2, 512]). size mismatch for action_distribution.std.bias: copying a param with shape torch.Size([3]) from checkpoint, the shape in current model is torch.Size([2]).
And I dont know how to fix it.So how can I train the navigation used in VinL???
Thanks!!!
Hi Simar!Your work is very great!!
When I tried to reproduce the code of "VinL",I can finish the locomotion part,including training and test.
But when I tried to use the navigation policy trained by myself,following the instruction in https://github.com/joannetruong/habitat-lab/blob/kin2dyn/README.md,it does not work.The report is followed:
RuntimeError: Error(s) in loading state_dict for PointNavResNetPolicy: size mismatch for net.prev_action_embedding.weight: copying a param with shape torch.Size([32, 3]) from checkpoint, the shape in current model is torch.Size([32, 2]). size mismatch for net.tgt_embeding.weight: copying a param with shape torch.Size([32, 4]) from checkpoint, the shape in current model is torch.Size([32, 3]). size mismatch for net.visual_encoder.compression.0.weight: copying a param with shape torch.Size([102, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([137, 256, 3, 3]). size mismatch for net.visual_encoder.compression.1.weight: copying a param with shape torch.Size([102]) from checkpoint, the shape in current model is torch.Size([137]). size mismatch for net.visual_encoder.compression.1.bias: copying a param with shape torch.Size([102]) from checkpoint, the shape in current model is torch.Size([137]). size mismatch for net.visual_fc.1.weight: copying a param with shape torch.Size([512, 2040]) from checkpoint, the shape in current model is torch.Size([512, 2055]). size mismatch for action_distribution.mu.weight: copying a param with shape torch.Size([3, 512]) from checkpoint, the shape in current model is torch.Size([2, 512]). size mismatch for action_distribution.mu.bias: copying a param with shape torch.Size([3]) from checkpoint, the shape in current model is torch.Size([2]). size mismatch for action_distribution.std.weight: copying a param with shape torch.Size([3, 512]) from checkpoint, the shape in current model is torch.Size([2, 512]). size mismatch for action_distribution.std.bias: copying a param with shape torch.Size([3]) from checkpoint, the shape in current model is torch.Size([2]).And I dont know how to fix it.So how can I train the navigation used in VinL???
Thanks!!!