Instructions to use alphachu-volleyball/alphachu-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- stable-baselines3
How to use alphachu-volleyball/alphachu-v1 with stable-baselines3:
from huggingface_sb3 import load_from_hub checkpoint = load_from_hub( repo_id="alphachu-volleyball/alphachu-v1", filename="{MODEL FILENAME}.zip", ) - Notebooks
- Google Colab
- Kaggle
alphachu-v1
First deployed model for world-tournament.
Source
- Experiment: 016_continue_015_extend, checkpoint_000015 (step ~111M)
- W&B run: https://wandb.ai/ootzk/alphachu-volleyball/runs/tv00fzku
- Training: curriculum crossplay, pre-duckll:5 unlock
Model config
| Field | Value |
|---|---|
| Side | player_1 |
| Action simplified | true |
| Obs simplified | false |
| Obs normalized | true |
| Size | 29.7 KB (ONNX) |
Win rates at checkpoint_000015 (Iter 15, 100 games each)
| Opponent | Win | Loss | Win rate |
|---|---|---|---|
| random | 100 | 0 | 100% |
| stone | 100 | 0 | 100% |
| builtin | 100 | 0 | 100% |
| duckll:0 | 100 | 0 | 100% |
| duckll:1 | 100 | 0 | 100% |
| duckll:2 | 94 | 6 | 94% |
| duckll:3 | 90 | 10 | 90% |
| duckll:7 | 85 | 15 | 85% |
Average pool win rate: 96.1%
Gameplay videos (5-point matches)
vs human
vs duckll:7
vs duckll:3
vs duckll:2
vs duckll:1
vs duckll:0
vs builtin
vs stone
vs random
- Downloads last month
- -