Spaces:
Sleeping
Sleeping
tasks file ai tasks stored and grader file stores ai reward or score
Browse files- graders.py +153 -0
- tasks.py +67 -0
graders.py
ADDED
|
@@ -0,0 +1,153 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# graders.py β reward functions and graders for all 3 tasks
|
| 2 |
+
|
| 3 |
+
|
| 4 |
+
# βββββββββββββββββββββββββββββ
|
| 5 |
+
# REWARD β called every single day
|
| 6 |
+
# βββββββββββββββββββββββββββββ
|
| 7 |
+
|
| 8 |
+
def compute_reward(soil_moisture, crop_health, water_action, rainfall):
|
| 9 |
+
"""
|
| 10 |
+
Returns a daily score between -1.0 and 1.0
|
| 11 |
+
"""
|
| 12 |
+
reward = 0.0
|
| 13 |
+
|
| 14 |
+
# Moisture check β ideal is 40% to 70%
|
| 15 |
+
if 40 <= soil_moisture <= 70:
|
| 16 |
+
reward += 0.5
|
| 17 |
+
elif 30 <= soil_moisture < 40 or 70 < soil_moisture <= 80:
|
| 18 |
+
reward += 0.2
|
| 19 |
+
else:
|
| 20 |
+
reward -= 0.3
|
| 21 |
+
|
| 22 |
+
# Crop health check
|
| 23 |
+
if crop_health >= 70:
|
| 24 |
+
reward += 0.3
|
| 25 |
+
elif crop_health >= 50:
|
| 26 |
+
reward += 0.1
|
| 27 |
+
else:
|
| 28 |
+
reward -= 0.2
|
| 29 |
+
|
| 30 |
+
# Penalize watering heavily when it's already raining a lot
|
| 31 |
+
if rainfall > 5 and water_action == 3:
|
| 32 |
+
reward -= 0.2
|
| 33 |
+
|
| 34 |
+
# Keep between -1.0 and 1.0
|
| 35 |
+
return round(max(-1.0, min(1.0, reward)), 2)
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
# βββββββββββββββββββββββββββββ
|
| 39 |
+
# TASK 1 GRADER β Easy (7 days)
|
| 40 |
+
# βββββββββββββββββββββββββββββ
|
| 41 |
+
|
| 42 |
+
def grade_task1(episode_history):
|
| 43 |
+
"""
|
| 44 |
+
Score = fraction of days where moisture was between 40β70%
|
| 45 |
+
Returns 0.0 to 1.0
|
| 46 |
+
"""
|
| 47 |
+
if not episode_history:
|
| 48 |
+
return 0.0
|
| 49 |
+
|
| 50 |
+
days_in_range = 0
|
| 51 |
+
for record in episode_history:
|
| 52 |
+
moisture = record.get("soil_moisture", 0)
|
| 53 |
+
if 40 <= moisture <= 70:
|
| 54 |
+
days_in_range += 1
|
| 55 |
+
|
| 56 |
+
score = days_in_range / len(episode_history)
|
| 57 |
+
return round(score, 2)
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
# βββββββββββββββββββββββββββββ
|
| 61 |
+
# TASK 2 GRADER β Medium (14 days)
|
| 62 |
+
# βββββββββββββββββββββββββββββ
|
| 63 |
+
|
| 64 |
+
def grade_task2(episode_history):
|
| 65 |
+
"""
|
| 66 |
+
Score based on daily rewards across 14 days with weather.
|
| 67 |
+
Returns 0.0 to 1.0
|
| 68 |
+
"""
|
| 69 |
+
if not episode_history:
|
| 70 |
+
return 0.0
|
| 71 |
+
|
| 72 |
+
total_reward = 0.0
|
| 73 |
+
max_possible = len(episode_history) * 0.8
|
| 74 |
+
|
| 75 |
+
for record in episode_history:
|
| 76 |
+
daily = compute_reward(
|
| 77 |
+
record.get("soil_moisture", 0),
|
| 78 |
+
record.get("crop_health", 0),
|
| 79 |
+
record.get("water_action", 0),
|
| 80 |
+
record.get("rainfall", 0)
|
| 81 |
+
)
|
| 82 |
+
total_reward += daily
|
| 83 |
+
|
| 84 |
+
score = total_reward / max_possible if max_possible > 0 else 0.0
|
| 85 |
+
return round(max(0.0, min(1.0, score)), 2)
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
# βββββββββββββββββββββββββββββ
|
| 89 |
+
# TASK 3 GRADER β Hard (30 days)
|
| 90 |
+
# βββββββββββββββββββββββββββββ
|
| 91 |
+
|
| 92 |
+
def grade_task3(episode_history):
|
| 93 |
+
"""
|
| 94 |
+
Score based on crop stage requirements over 30 days.
|
| 95 |
+
Seedling (days 1-10), Growing (days 11-20), Mature (days 21-30)
|
| 96 |
+
Returns 0.0 to 1.0
|
| 97 |
+
"""
|
| 98 |
+
if not episode_history:
|
| 99 |
+
return 0.0
|
| 100 |
+
|
| 101 |
+
day_scores = []
|
| 102 |
+
|
| 103 |
+
for record in episode_history:
|
| 104 |
+
day = record.get("day", 1)
|
| 105 |
+
moisture = record.get("soil_moisture", 0)
|
| 106 |
+
crop_health = record.get("crop_health", 0)
|
| 107 |
+
|
| 108 |
+
# Each stage needs different moisture
|
| 109 |
+
if day <= 10: # Seedling
|
| 110 |
+
ideal = 50 <= moisture <= 75
|
| 111 |
+
elif day <= 20: # Growing
|
| 112 |
+
ideal = 45 <= moisture <= 70
|
| 113 |
+
else: # Mature
|
| 114 |
+
ideal = 35 <= moisture <= 60
|
| 115 |
+
|
| 116 |
+
if ideal and crop_health >= 70:
|
| 117 |
+
day_scores.append(1.0)
|
| 118 |
+
elif ideal or crop_health >= 60:
|
| 119 |
+
day_scores.append(0.5)
|
| 120 |
+
else:
|
| 121 |
+
day_scores.append(0.0)
|
| 122 |
+
|
| 123 |
+
score = sum(day_scores) / len(day_scores)
|
| 124 |
+
return round(score, 2)
|
| 125 |
+
|
| 126 |
+
|
| 127 |
+
# βββββββββββββββββββββββββββββ
|
| 128 |
+
# TEST β run this file to verify
|
| 129 |
+
# βββββββββββββββββββββββββββββ
|
| 130 |
+
|
| 131 |
+
if __name__ == "__main__":
|
| 132 |
+
# Fake data to test all 3 graders
|
| 133 |
+
fake_7 = [
|
| 134 |
+
{"day": i, "soil_moisture": 55, "crop_health": 75,
|
| 135 |
+
"water_action": 2, "rainfall": 0}
|
| 136 |
+
for i in range(1, 8)
|
| 137 |
+
]
|
| 138 |
+
fake_14 = [
|
| 139 |
+
{"day": i, "soil_moisture": 50 + i, "crop_health": 72,
|
| 140 |
+
"water_action": 1, "rainfall": i % 3}
|
| 141 |
+
for i in range(1, 15)
|
| 142 |
+
]
|
| 143 |
+
fake_30 = [
|
| 144 |
+
{"day": i, "soil_moisture": 55, "crop_health": 70 + (i % 10),
|
| 145 |
+
"water_action": 2, "rainfall": i % 5}
|
| 146 |
+
for i in range(1, 31)
|
| 147 |
+
]
|
| 148 |
+
|
| 149 |
+
print("Task 1 Score:", grade_task1(fake_7))
|
| 150 |
+
print("Task 2 Score:", grade_task2(fake_14))
|
| 151 |
+
print("Task 3 Score:", grade_task3(fake_30))
|
| 152 |
+
|
| 153 |
+
print("\nSample daily reward:", compute_reward(55, 78, 2, 0))
|
tasks.py
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# tasks.py β defines all 3 tasks and their action schema
|
| 2 |
+
|
| 3 |
+
TASKS = [
|
| 4 |
+
{
|
| 5 |
+
"task_id": 1,
|
| 6 |
+
"name": "Basic Moisture Control",
|
| 7 |
+
"difficulty": "easy",
|
| 8 |
+
"description": "Keep soil moisture between 40β70% for 7 days.",
|
| 9 |
+
"max_steps": 7,
|
| 10 |
+
"goal": "moisture_in_range"
|
| 11 |
+
},
|
| 12 |
+
{
|
| 13 |
+
"task_id": 2,
|
| 14 |
+
"name": "Weather Adaptive Irrigation",
|
| 15 |
+
"difficulty": "medium",
|
| 16 |
+
"description": "Manage irrigation across 14 days with changing weather.",
|
| 17 |
+
"max_steps": 14,
|
| 18 |
+
"goal": "handle_weather"
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"task_id": 3,
|
| 22 |
+
"name": "Full Season Management",
|
| 23 |
+
"difficulty": "hard",
|
| 24 |
+
"description": "Manage a full 30-day crop season across Seedling, Growing, and Mature stages.",
|
| 25 |
+
"max_steps": 30,
|
| 26 |
+
"goal": "full_season"
|
| 27 |
+
}
|
| 28 |
+
]
|
| 29 |
+
|
| 30 |
+
ACTION_SCHEMA = {
|
| 31 |
+
"type": "discrete",
|
| 32 |
+
"n": 4,
|
| 33 |
+
"actions": {
|
| 34 |
+
0: "no_water",
|
| 35 |
+
1: "low_water",
|
| 36 |
+
2: "medium_water",
|
| 37 |
+
3: "high_water"
|
| 38 |
+
},
|
| 39 |
+
"description": "Choose how much water to give the crop each day."
|
| 40 |
+
}
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def get_all_tasks():
|
| 44 |
+
"""Returns the list of all tasks with action schema."""
|
| 45 |
+
return {
|
| 46 |
+
"tasks": TASKS,
|
| 47 |
+
"action_schema": ACTION_SCHEMA
|
| 48 |
+
}
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
def get_task_by_id(task_id):
|
| 52 |
+
"""Returns a single task by its ID."""
|
| 53 |
+
for task in TASKS:
|
| 54 |
+
if task["task_id"] == task_id:
|
| 55 |
+
return task
|
| 56 |
+
return None
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
# Quick test
|
| 60 |
+
if __name__ == "__main__":
|
| 61 |
+
print("All tasks:")
|
| 62 |
+
result = get_all_tasks()
|
| 63 |
+
for task in result["tasks"]:
|
| 64 |
+
print(f" Task {task['task_id']}: {task['name']} ({task['difficulty']})")
|
| 65 |
+
print("\nAction schema:")
|
| 66 |
+
for k, v in result["action_schema"]["actions"].items():
|
| 67 |
+
print(f" {k} = {v}")
|