sac-lili

Files:

• softlearning/algorithms/sac.py: implements SAC and latent model
• softlearning/replay_pools/multitask_replay_pool.py: implements replay pool separated by task
• softlearning/samplers/simple_sampler.py: encodes trajectory from previous task and samples from the environment with policy conditioned on encoding

Setup:

cd softlearning
conda env create -f environment.yml
conda activate sac_lili
pip install -e .

The hyperparameters for LILI are in examples/development/variants.py.