Update README.md
Browse files
README.md
CHANGED
|
@@ -1,15 +1,55 @@
|
|
| 1 |
---
|
| 2 |
title: Ai Training Simulator
|
| 3 |
-
emoji:
|
| 4 |
colorFrom: blue
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.16.0
|
| 8 |
python_version: '3.13'
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
| 11 |
license: mit
|
| 12 |
-
short_description: Idle/
|
| 13 |
---
|
| 14 |
|
| 15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
title: Ai Training Simulator
|
| 3 |
+
emoji: 🧠
|
| 4 |
colorFrom: blue
|
| 5 |
+
colorTo: green
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.16.0
|
| 8 |
python_version: '3.13'
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
| 11 |
license: mit
|
| 12 |
+
short_description: Idle/strategy game
|
| 13 |
---
|
| 14 |
|
| 15 |
+
# AI Training Simulator
|
| 16 |
+
|
| 17 |
+
**Build Small Hackathon 2026 — Chapter Two (Adventure in Thousand Token Wood)**
|
| 18 |
+
|
| 19 |
+
A browser strategy/idle game where you manage a fake AI startup.
|
| 20 |
+
|
| 21 |
+
## How it works
|
| 22 |
+
|
| 23 |
+
- Start with **$100**
|
| 24 |
+
- Buy upgrades (model size, dataset size/quality, architecture, hardware)
|
| 25 |
+
- Click **Train Model** — cost and reward are simulated based on your upgrade combination
|
| 26 |
+
- The chat interface uses **SmolLM2-360M-Instruct running entirely in your browser** via transformers.js
|
| 27 |
+
- Low quality → high temperature → incoherent word salad
|
| 28 |
+
- Upgrade your stack → temperature drops → coherent output
|
| 29 |
+
- Unlock **Cloud Inference** for recurring revenue
|
| 30 |
+
|
| 31 |
+
## Badges targeted
|
| 32 |
+
- 🔌 Off the Grid — no cloud APIs, model runs client-side
|
| 33 |
+
- 🎨 Off-Brand — fully custom UI (no default Gradio look)
|
| 34 |
+
|
| 35 |
+
## Files
|
| 36 |
+
|
| 37 |
+
| File | Purpose |
|
| 38 |
+
|------|---------|
|
| 39 |
+
| `app.py` | Gradio app + HTML game UI + action handler |
|
| 40 |
+
| `state.py` | Pure Python game logic (train, upgrade, cloud) |
|
| 41 |
+
| `game_config.py` | **All balance numbers** — edit this to tune the game |
|
| 42 |
+
| `requirements.txt` | Only `gradio` needed (model runs in browser) |
|
| 43 |
+
|
| 44 |
+
## Customising / balancing
|
| 45 |
+
|
| 46 |
+
Everything tunable is in `game_config.py`:
|
| 47 |
+
- `STARTING_MONEY` — how much the player starts with
|
| 48 |
+
- `UPGRADES` — levels, costs, quality scores per upgrade
|
| 49 |
+
- `BASE_TRAIN_COST` / multiplier tables — training run economics
|
| 50 |
+
- `BASE_REWARD` + `QUALITY_EXPONENT` — reward curve shape
|
| 51 |
+
- `REWARD_NOISE_PCT` — variance per run
|
| 52 |
+
- `CLOUD_*` constants — cloud inference economics
|
| 53 |
+
- `CHAT_TEMP_*` — how temperature maps to model quality
|
| 54 |
+
|
| 55 |
+
Author: Martico2432
|