Autori: Luca Gregori & Alessandro Wood
Corso: Machine Learning
Si veda il notebook Relazione Finale
- Installazione driver Nvidia (proprietari) (prerequisiti per cuda):
sudo apt install nvidia-driver-460
- Installazione Cuda (versione 11.2):
sudo apt install nvidia-cuda-toolkit
-
Installazione nvidia cuDNN dal seguente link (richiede registrazione al programma nvidia developer)
cuDNN Runtime Library for Ubuntu20.04 x86_64 (Deb) è il pacchetto da installare -
Esportazioni variabili d’ambiente:
echo 'export LD_LIBRARY_PATH=/usr/lib/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/lib/cuda/include:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc
- Per verificare l’installazione di cuda
nvcc -V
-
Installazione di Tensorflow/PyTorch tramite pip
-
Installazione ViZDoom si segua la guida quick start per python: VizDoom.
Arnold : @inproceedings{chaplot2017arnold, title={Arnold: An Autonomous Agent to Play FPS Games.}, author={Chaplot, Devendra Singh and Lample, Guillaume}, booktitle={Proceedings of AAAI}, year={2017}, Note={Best Demo award} }
ViZDoom: @article{wydmuch2018vizdoom, title={ViZDoom Competitions: Playing Doom from Pixels}, author={Wydmuch, Marek and Kempka, Micha{\l} and Ja{'s}kowski, Wojciech}, journal={IEEE Transactions on Games}, year={2018}, publisher={IEEE} Rarity of Events: @article{roe, title={Automated Curriculum Learning by Rewarding Temporally Rare Events}, author={Niels Justesen, Sebastian Risi}, year={2018}
Automated Curriculum Learning by Rewarding Temporally Rare Events | Niels Justesen & Sebastian Risi | link
Human-level control through deep reinforcement learning | DeepMind | 26 february 2015 | vol 518 | Nature | link