Breadway/novacana-engine
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
Novacana card game engine and ML training
14 commits 3 branches 0 tags 197 MiB
  • Rust 90.6%
  • Python 7.8%
  • Shell 1%
  • Batchfile 0.3%
  • PowerShell 0.3%
Find a file
Exact
breadway f51fcf714a Stabilize Astra-7 BC common-offset gradients
2026-10-05 05:25:40 +08:00
brothers-pc Ignore generated artifacts and remove tracked outputs 2026-10-04 21:40:58 +08:00
client Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
docs Stabilize Astra-7 BC common-offset gradients 2026-10-05 05:25:40 +08:00
engine Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
ml Ignore generated artifacts and remove tracked outputs 2026-10-04 21:40:58 +08:00
ml7 Stabilize Astra-7 BC common-offset gradients 2026-10-05 05:25:40 +08:00
server Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
src Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
tools Ignore generated artifacts and remove tracked outputs 2026-10-04 21:40:58 +08:00
.gitignore Ignore generated artifacts and remove tracked outputs 2026-10-04 21:40:58 +08:00
ASTRA7.md Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
Cargo.lock Add opt-in two-device Astra-7 behavior cloning 2026-10-05 04:19:36 +08:00
Cargo.toml Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
GPU_RENTAL.md Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
LICENSE Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
Overview.md Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
README.md Expand Astra-7 training, GPU support, and monitoring 2026-10-04 21:33:49 +08:00
STATE.md can't be bothered writing a commit message 2026-10-05 03:09:06 +08:00

README.md

Novacana

Self-play ML training engine for a card/board game AI (working name ASTRA), built as a Rust workspace: game simulation, self-play data generation, and model training/eval.

Cargo workspace members:

  • engine - game simulation / self-play
  • client, server - play/serve the game
  • ml, ml7 - model training and evaluation

See ASTRA7.md for the current implementation plan and GPU_RENTAL.md for the training infrastructure setup.

Building

cargo build --release