Python=3.9
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
pip install -U diffusers accelerate transformers
pip install einops matplotlib wandb
pip install dm-reverb[tensorflow] tensorflow-datasets rlds
pip install "pydantic>=2.0" --upgrade
git clone https://github.com/huggingface/diffusers
cd diffusers
pip install -e .
pip install accelerate transformers datasets
pip install einops matplotlib wandb
pip install "pydantic>=2.0" --upgrade
pip install dm-reverb[tensorflow] tensorflow-datasets rlds
pip install xformers
This section explains using CALVIN dataset to train our model.
To generate optical flow and save, run following script:
cd src/dataset
python generate_flow.pyTo visualize generation data, run
python test/dataset/test_generated_flow.pyAdditional code for visualization and sanity checks on CALVIN optical flow generation
are found in src/dataset/calvin.py.
python src/train.py --output-dir test_0XX