jan/rusty_marvin
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

New evaluator that uses only 133 MB of memory and improves performance of 6- and 7-card hand evaluations

Browse source
This commit is contained in:
Jan-Peter von Hunnius 2025-10-11 20:30:50 +02:00
parent 696bf46555
commit 7819b1ee27
18 changed files with 4367 additions and 7701 deletions
Download patch file Download diff file Expand all files Collapse all files

18
docs/VISION.md
View file

@ -11,12 +11,20 @@ Create a **comprehensive poker AI ecosystem** with two main components:
## Core Architecture Components
### 1. Simulation Testbed
### 1. Advanced Evaluation Engine
- **Configurable bots** (Rule-based, Nash EQ, Statistical, Hand-reading, Reinforcement Learning)
- **Mass parallel simulations** with GPU acceleration (e.g., for turn and river rollouts)
- **TOML-driven configuration** for all parameters of bots and simulations
- **Comprehensive result analysis** and bot evolution
- **Perfect Hash Jump Tables**: Revolutionary memory-efficient evaluation (~130MB vs 625MB)
- **Suit Canonicalization**: Advanced isomorphic hand reduction for optimal performance
- **Bottom-up Trie Construction**: Cache-optimized memory layout for CPU efficiency
- **Zero Breaking Changes**: Full backward compatibility with existing evaluation APIs
### 2. Simulation Testbed
- **Configurable bots** (Rule-based, Nash EQ, Statistical, Hand-reading, Reinforcement Learning)
- **Mass parallel simulations** with GPU acceleration (e.g., for turn and river rollouts)
- **Memory-efficient evaluation**: 79% memory reduction enables larger-scale simulations
- **TOML-driven configuration** for all parameters of bots and simulations
- **Comprehensive result analysis** and bot evolution
### 2. Live Action Server