JdeRobot team allignment & inference rlstudio module created (month 17)
First of all. We finally got a powerfull personal computer. This is relevant not just because it will enable us to use more resources-demanding algorithms from now on, but also because the final allignment within the team was possible. Lets dig into it:
Computer Alienware Aurora R10 requirements:
- AMD Ryzen 9 5900X 3.7GHz
- NVIDIA GeForce RTX 3080 10GB GDDR6X
- 32GB DDR4-3400 RAM1TB Solid State Drive
- 1TB Solid State Drive
Team alignment through the f1 rlstudio problem:
- Due to the lack of a goot computer, the problem was not performing well. The camera image was a little bit lagged so the actions learnt by the f1 and the actions accomplished due to the algorithm decision were not working as expected.
- Now, using the uploaded config.yaml. The robot really learn what it has to learn and complete the lap successfully.
That said, the tasks for this month were:
- Merging into the main branch the robot-mesh and mountain car code and documentation
- Creating an inference module to enable rl-studio to provide a way of using the trained brains in an inference mode out of the rl-studio environment (e.g taking the brain out of the tool and using in real world in inference mode)
So, the outcomes for this months are listed below: