Aditya Prakash

Can you say more about the model training with 20% util? Which model did you train? what batch size did you use? how much GPU memory was occupied?

The GPU-util for NEAT is expected to be lower than other models since the dataloader for neat involves a lot more computation but I can't tell by what factor. If...

1. You can assign the desired location to save data to the `SAVE_PATH` variable in [./leaderboard/scripts/run_evaluation.sh](https://github.com/autonomousvision/neat/blob/main/leaderboard/scripts/run_evaluation.sh#L28) to generate data. 2. The data is generated only once for all the models...

The decoder predicts the current timestep and 4 future timesteps and we require the future timesteps predictions for controller.

Semantic loss is also crucial for training our model. You can see in Table 3 our [supplementary material](http://www.cvlibs.net/publications/Chitta2021ICCV_supplementary.pdf) that there is a significant drop in performance when we remove the...

The main goal is to learn pouring policies that can handle variation in the geometry of the cup/glass. For that, I need to run simulation with different objects. So, I...

> Great detective work! I have been trying to figure out why I keep running out of GPU memory, and this seems to be the problem. I get through about...

> > > Great detective work! I have been trying to figure out why I keep running out of GPU memory, and this seems to be the problem. I get...

We trained and evaluated on 0.8.4. For CARLA 0.9.x you need to regenerate data and make slight modifications in the architecture to account for 6 navigational commands in 0.9.x compared...