Goal Representations for Instruction Following (GRIF)

This is the code repository for the paper Goal Representations for Instruction Following: A Semi-Supervised Language Interface to Control [arXiv, website].

Based on dibyaghosh/jaxrl_minimal.

Environment

For GPU:

conda env create -f environment_cuda.yml

For TPU:

conda env create -f environment_tpu.yml

See the Jax Github page for more details on installing Jax.

Running

To train GRIF, run

bash experiments/scripts/launch_bridge.sh GRIF