# transfuser

**Repository Path**: zhaoyewei/transfuser

## Basic Information

- **Project Name**: transfuser
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: leaderboard_submission
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-07-09
- **Last Updated**: 2021-07-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

## CARLA Leaderboard Submission

Read the instructions on the [leaderboard](https://leaderboard.carla.org/get_started/).

### Building docker image

Add the following paths to your ```~/.bashrc```
```
export CARLA_ROOT=<path_to_carla_root>
export SCENARIO_RUNNER_ROOT=<path_to_scenario_runner_in_this_repo>
export LEADERBOARD_ROOT=<path_to_leaderboard_in_this_repo>
export PYTHONPATH="${CARLA_ROOT}/PythonAPI/carla/":"${SCENARIO_RUNNER_ROOT}":"${LEADERBOARD_ROOT}":${PYTHONPATH}
```

Edit the contents of ```leaderboard/scripts/Dockerfile.master``` to specify the required dependencies, agent code and model checkpoints. Add all the required information in the area delimited by the tags ```BEGINNING OF USER COMMANDS``` and ```END OF USER COMMANDS```. The current Dockerfile works for all the models in this repository.

Specify a name for the docker image in ```leaderboard/scripts/make_docker.sh``` and run:
```
leaderboard/scripts/make_docker.sh
```

Refer to the Transfuser example for the directory structure and where to include the code and checkpoints.

### Testing the docker image locally

Spin up a CARLA server:
```
SDL_VIDEODRIVER=offscreen SDL_HINT_CUDA_DEVICE=0 ./CarlaUE4.sh -world-port=2000 -opengl
```

Run the docker container:  
Docker 19:  
```
docker run -it --rm --net=host --gpus '"device=0"' -e PORT=2000 <docker_image> ./leaderboard/scripts/run_evaluation.sh
```

Docker 18:  
```
docker run -it --rm -p 2000-2002:2000-2002 --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0 -e PORT=2000 <docker_image> ./leaderboard/scripts/run_evaluation.sh
```
If the docker container doesn't start properly, add another environment variable ```SDL_AUDIODRIVER=dsp```.

### Submitting docker image to the leaderboard

Register on [AlphaDriver](https://app.alphadrive.ai/), create a team and apply to the CARLA Leaderboard.

Install AlphaDrive cli:
```
curl http://dist.alphadrive.ai/install-ubuntu.sh | sh -
```

Login to alphadrive and submit the docker image:
```
alpha login
alpha benchmark:submit --split <2/3> <docker_image>
```
Use ```split 2``` for MAP track and ```split 3``` for SENSORS track.