Pretrained model not accessible

[nweir127]

Hello, I'm having trouble loading your released hierarchical model checkpoint-- it seems to be trying to access a class from pytorch-a2c-ppo-acktr that doesn't exist.

>>> import torch
>>> torch.load('trained_models/our_model.pt')
/anaconda/envs/mazebase3.6/lib/python3.6/site-packages/torch/serialization.py:434: SourceChangeWarning: source code of class 'pytorchppo.a2c_ppo_acktr.model.Policy' has changed. you can retrieve the original source code by accessing the object's source attribute or set `torch.nn.Module.dump_patches = True` and use the patch tool to revert the changes.
  warnings.warn(msg, SourceChangeWarning)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/anaconda/envs/mazebase3.6/lib/python3.6/site-packages/torch/serialization.py", line 367, in load
    return _load(f, map_location, pickle_module)
  File "/anaconda/envs/mazebase3.6/lib/python3.6/site-packages/torch/serialization.py", line 538, in _load
    result = unpickler.load()
AttributeError: Can't get attribute 'MazeBaseBase' on <module 'pytorchppo.a2c_ppo_acktr.model' from '/path/to/ask-your-humans/mazebase-training/pytorchppo/a2c_ppo_acktr/model.py'>

Any advice would be greatly appreciated.

[valeriechen]

You're right, sorry about that - I'll update the code to fix this issue this weekend!

[nweir127]

Thanks very much!

Also FYI-- I wasn't able to run pip install -r requirements in mazebase-training and had to install various packages using conda install instead. I am assuming your environment is python 3.6?

[valeriechen]

Hi! Here's how you can run our model:

• Download dataset split.zip to a folder called data
• Download our_model
• Run python enjoy.py --load-dir our_model. See updated instructions about how to comment out things in the readme if you want to run different models (apologies in advance for the formatting issues).

Happy to help with any other questions you might have!