The Img_encoder pre-trained model incorrect!!!

[jishnujayakumar]

Hi @bareblackfoot,

As mentioned in #6 and #7, Img_encoder link in visualize not working. The Img_encoder in the data/graph is not loading properly—model missing keys. Please share the correct one so that build_tgsm.ipynb can be run.

[bareblackfoot]

fixed :)
Please check it!

[07hokage]

Hi @bareblackfoot

Based on the requirements given on the README, we tried installing and running build_tgsm.ipynb but got the following error while loading the Img_encoder.pth. We have tried on multiple systems still the error persists.

/home/experiments/TopologicalSemanticGraphMemory
/home/user/anaconda3/envs/tsgm/lib/python3.7/site-packages/habitat
Traceback (most recent call last):
  File "test_build_tsgm.py", line 59, in <module>
    img_encoder.load_state_dict(state_dict)
  File "/home/user/anaconda3/envs/tsgm/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1483, in load_state_dict
    self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for ResNet:
        size mismatch for conv1.weight: copying a param with shape torch.Size([64, 3, 7, 7]) from checkpoint, the shape in current model is torch.Size([64, 4, 7, 7]).