Mila has many people working on reinforcement learning research. RL research often has a high bar for entry because of the system-level requirements to access and use hardware, environments and RL algorithms. We should pool our resources and skills to make it easier for students to perform RL research and advance our methods to tackle new research areas.

Topics for discussion

  1. Using GPUs properly for RL
  2. Multi-Processing for RL

Logging your experiments

  1. Accessing RL environments

Distributed computing across computers

  1. Making your code mobile (easy to run across different compute clusters)


Time Speaker Topic
1pm-1:30 pm Gev (Aim) Experiment tracking
1:30 - 2pm Max Schwarzer Jax for the PyTorch Native
2 - 2:30pm Jacob Buckman Maximizing GPU throughput in RL
2:30 - 3pm Melissa/Johanna Sim2Real Robotics mini-tutorial
3:30 - 4pm Jesse Farebrother Code Mobility
4 - 4:30pm Niki Howe Myriad: a real-world testbed to bridge trajectory optimization and deep learning
4:30 - 5pm Darhsan Patil RLHive