Skip to content

Latest commit

 

History

History
44 lines (34 loc) · 2.1 KB

INSTALL.md

File metadata and controls

44 lines (34 loc) · 2.1 KB
Raw

Installation

This doc provides instructions to get started.

Install CARLA

Install dependencies

  • First, inside the repo, create a dedicated conda environment. Refer here if you do not have conda.
conda env create -f environment.yaml
  • Inside the conda environment, install the CARLA PythonAPI easy_install [PATH TO CARLA EGG]. Refer to this link if you are confused at this step.
  • Install PyTorch
  • Install torch-scatter based on your CUDA and PyTorch versions.
  • Setup wandb

Configure environment variables

Note: the following instructions only apply to Linux. For Windows follow these instructions instead.

Set the following environmental variables to your conda environment. Refer here for instructions to do so.

#!/bin/bash

export CARLA_ROOT=[LINK TO YOUR CARLA FOLDER]
export LEADERBOARD_ROOT=[LINK TO LAV REPO]/leaderboard
export SCENARIO_RUNNER_ROOT=[LINK TO LAV REPO]/scenario_runner
export PYTHONPATH="${CARLA_ROOT}/PythonAPI/carla/":"${SCENARIO_RUNNER_ROOT}":"${LEADERBOARD_ROOT}"
export TEAM_AGENT=[LINK TO LAV REPO]/team_code_v2/lav_agent.py
export TEAM_CONFIG=[LINK TO LAV REPO]/team_code_v2/config.yaml

export SCENARIOS=${LEADERBOARD_ROOT}/data/all_towns_traffic_scenarios_public.json
export REPETITIONS=1
export CHECKPOINT_ENDPOINT=results.json
export DEBUG_CHALLENGE=0
export CHALLENGE_TRACK_CODENAME=SENSORS

Now, you can treat this conda environment the dedicated one (by defauld named LAV-env).