hbofz/NeSLE

Windows WDDM caps native PPO throughput at ~3k env-steps/s 19 days ago

bug help wanted

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Title screen to gameplay transition stalls on PPU timing 19 days ago

bug help wanted

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Renderer: residual lag-frame artifacts drop the status bar 19 days ago

bug help wanted

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Mapper support beyond NROM (MMC1, MMC3) 19 days ago

enhancement help wanted

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Regenerate the stale GTX 1050 Ti benchmark table 19 days ago

documentation good first issue

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Fold the CUDA extension build into setup.py 19 days ago

good first issue

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Measure a multi-core CPU baseline for a fair speedup comparison 19 days ago

documentation good first issue

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Run the C++ tests on Windows and surface them through pytest 19 days ago

good first issue

hbofz/NeSLE

1

GPU-native NES emulator + RL stack: thousands of parallel NES envs on one CUDA GPU, training PPO agents on Super Mario Bros.

Python

#cuda#emulator#gpu#gymnasium#machine-learning#nes#ppo#pybind11#reinforcement-learning#super-mario-bros

Open Issues Need Help