Spaces:
Running on Zero
Running on Zero
File size: 6,264 Bytes
e9fc2fc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 | """
Test the game generation pipeline with llama.cpp and NVIDIA Nemotron 3 Nano 4B GGUF.
Run this script to:
1. Test model availability (llama-cpp-python)
2. Demonstrate GGUF model loading from HuggingFace
3. Generate games with the Nemotron model
4. Fall back to mock generation if model unavailable
5. Validate all outputs against schema
"""
import json
from pathlib import Path
from app.services.retrieval import load_games_dataset, normalize_game_record, retrieve_examples
from app.services.generator import generate_game, build_generation_prompt, NEMOTRON_MODEL_ID, NEMOTRON_GGUF_FILE
from app.services.schema_validator import validate_game_schema
def check_llama_cpp_availability():
"""Check if llama-cpp-python is installed."""
print("\n" + "=" * 80)
print("CHECKING ENVIRONMENT")
print("=" * 80)
try:
import llama_cpp
print(f"✓ llama-cpp-python is installed")
print(f" Version: {llama_cpp.__version__ if hasattr(llama_cpp, '__version__') else 'unknown'}")
return True
except ImportError:
print("✗ llama-cpp-python not found")
print(" Install with: pip install llama-cpp-python")
print(" Or for GPU support: pip install llama-cpp-python[cuda]")
return False
def main():
print("\n" + "=" * 80)
print("PHASE 2, TASK 6: GAME GENERATION WITH NEMOTRON 3 NANO 4B GGUF")
print("=" * 80)
# Check environment
llama_cpp_available = check_llama_cpp_availability()
print(f"\nModel Configuration:")
print(f" Repository: {NEMOTRON_MODEL_ID}")
print(f" File: {NEMOTRON_GGUF_FILE}")
print(f" Runtime: llama.cpp (GGUF quantized)")
print(f" Benefits:")
print(f" • GGUF quantization: 4-bit, memory efficient")
print(f" • llama.cpp: Fast CPU/GPU inference")
print(f" • Hackathon bonus: Extra credit for llama.cpp runtime")
# Load dataset
print("\n1. Loading dataset...")
raw_records = load_games_dataset("app/data/games_dataset.json")
normalized_records = [normalize_game_record(r) for r in raw_records]
print(f"✓ Loaded {len(normalized_records)} records")
# Test cases
test_configs = [
{
"name": "Scavenger Hunt - Adults - Medium",
"config": {
"game_type": "scavenger_hunt",
"city": "Paris",
"area": "Le Marais",
"location_type": "mixed",
"duration_minutes": 60,
"num_players": 4,
"difficulty": "medium",
"age_group": "adults"
}
},
{
"name": "Hide & Seek - Kids - Easy",
"config": {
"game_type": "hide_and_seek",
"city": "Paris",
"area": "Parc des Buttes-Chaumont",
"location_type": "park",
"duration_minutes": 45,
"num_players": 5,
"difficulty": "easy",
"age_group": "kids"
}
}
]
# Test generation
results = []
for test in test_configs:
print("\n" + "=" * 80)
print(f"TEST: {test['name']}")
print("=" * 80)
config = test['config']
# Retrieve similar games
print("\n2. Retrieving similar games...")
retrieved = retrieve_examples(config, normalized_records, k=3)
print(f"✓ Retrieved {len(retrieved)} examples")
# Build prompt
print("\n3. Building generation prompt...")
prompt = build_generation_prompt(config, retrieved)
print(f"✓ Prompt ready ({len(prompt)} chars)")
# Generate game
print("\n4. Generating game...")
print(f" (Using: NVIDIA Nemotron 3 Nano 4B GGUF via llama.cpp)")
try:
game = generate_game(config, retrieved)
print(f"✓ Game generated: {game['game_id']}")
# Validate
print("\n5. Validating against schema...")
is_valid, errors = validate_game_schema(game)
if is_valid:
print("✓ Game VALID against schema")
else:
print(f"✗ Validation errors: {len(errors)}")
# Display summary
print("\n6. Game Summary:")
print(f" Title: {game['title']}")
print(f" Area: {game['setup']['area']}")
print(f" Duration: {game['setup']['duration_minutes']} min | Players: {game['setup']['num_players']}")
print(f" Tasks: {len(game['tasks'])} | Rules: {len(game['rules'])}")
print(f" Tone: {game['story_seed']['tone']}")
results.append({
'name': test['name'],
'valid': is_valid,
'game_id': game['game_id']
})
except Exception as e:
print(f"✗ Generation failed: {e}")
results.append({
'name': test['name'],
'valid': False,
'game_id': None
})
# Summary
print("\n" + "=" * 80)
print("TEST SUMMARY")
print("=" * 80)
passed = sum(1 for r in results if r['valid'])
print(f"\nResults: {passed}/{len(results)} tests passed")
for result in results:
status = "✓ PASS" if result['valid'] else "✗ FAIL"
print(f"{status}: {result['name']}")
if result['game_id']:
print(f" {result['game_id']}")
print("\n" + "=" * 80)
print("NOTES")
print("=" * 80)
if not llama_cpp_available:
print("⚠ llama-cpp-python not installed - using mock generation")
print(" For actual model-based generation, install:")
print(" pip install llama-cpp-python[cuda] # For GPU")
print(" or")
print(" pip install llama-cpp-python # For CPU")
else:
print("✓ llama-cpp-python available")
print("✓ NVIDIA Nemotron 3 Nano 4B GGUF ready for download from HuggingFace")
print("✓ Hackathon extra credit: llama.cpp runtime ✓")
print("\n" + "=" * 80)
if __name__ == "__main__":
main()
|