Run High-Throughput Reinforcement Learning Training with End-to-End FP8 Precision

Develop an end-to-end FP8 training suite for Reinforcement Learning algorithms. Currently, FP8 is well-supported for inference, but training for RL agents using FP8 is a missing piece in most OSS stacks.

Suggested repo: fp8-rl

"Speed up your RL training loop by 3x with native FP8 support."

Estimated effort: 150h