Reproduce the paper: CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks
The original code: https://github.com/mees/calvin
To run the code, please download the dataset (provided in the above github repo):
wget http://calvin.cs.uni-freiburg.de/dataset/calvin_debug_dataset.zip
Please put the dataset at the same directory as train.py.
Or modify the address in Dataset.py:
Then, run train.py, and the result will be recorded in calvin.log.
The result I get is recorded in result.log.