sayed-bot-model-v2

Decision Transformer for Sayed Bot — Minecraft bot AI

Model

• Architecture: Transformer Encoder (6 layers, 8 heads, 512 dim)
• Input: [batch, 20, 15] — آخر 20 خطوة
• Output: [batch, 10] — logits لكل action
• Features: health, food, mobs_nearby, danger_mobs, ores_nearby, items_nearby, has_weapon, has_pickaxe, has_axe, inventory_pct, time_of_day, is_in_danger, is_on_lava, player_nearby, bot_count
• Actions: IDLE, FIGHT, FLEE, MINE, FARM, COLLECT, EXPLORE, CRAFT, CHAT, FOLLOW
• Accuracy: 33.23%
• Size: ~50 MB (INT8)
• Trained: 2026-06-25T11:38:05Z

Usage (ONNX Runtime)

import onnxruntime as ort
import numpy as np

session = ort.InferenceSession('bot_model.onnx')
input_name = session.get_inputs()[0].name

sequence = np.zeros((1, 20, 15), dtype=np.float32)
# املأ الـ sequence بـ 20 خطوة سابقة
result = session.run(None, {input_name: sequence})
action_id = result[0].argmax()

Files

• model.onnx
• config.json