|
|
|
|
|
""" |
|
|
Test script to verify the complete AlphaEvolve setup works |
|
|
""" |
|
|
|
|
|
import sys |
|
|
import json |
|
|
from pathlib import Path |
|
|
|
|
|
def test_evaluator(): |
|
|
"""Test the evaluator with a simple program""" |
|
|
|
|
|
print("π§ͺ Testing evaluator...") |
|
|
|
|
|
|
|
|
test_program = ''' |
|
|
def optimize_attention(): |
|
|
return { |
|
|
'tile_size_m': 32, |
|
|
'tile_size_n': 64, |
|
|
'vectorization': 'none', |
|
|
'unroll_factor': 2, |
|
|
'loop_interchange': False, |
|
|
'fusion_strategy': 'none', |
|
|
'use_shared_memory': False, |
|
|
'optimize_for_latency': True, |
|
|
'enable_blocking': False, |
|
|
'enable_recomputation': False, |
|
|
'optimization_strategy': 'alphaevolve_test', |
|
|
'target_speedup': 1.32, |
|
|
} |
|
|
''' |
|
|
|
|
|
try: |
|
|
|
|
|
sys.path.insert(0, '.') |
|
|
from evaluator import evaluate_program |
|
|
|
|
|
print("β
Evaluator imported successfully") |
|
|
|
|
|
|
|
|
result = evaluate_program(test_program) |
|
|
|
|
|
if 'error' in result: |
|
|
print(f"π Evaluation result: error={result['error']:.3f}") |
|
|
if 'speedup' in result: |
|
|
print(f"π Speedup: {result['speedup']:.3f}x") |
|
|
if 'mlir_source' in result: |
|
|
print(f"π MLIR source: {result['mlir_source']}") |
|
|
|
|
|
if result['error'] < 1000: |
|
|
print("β
Evaluator works!") |
|
|
return True |
|
|
else: |
|
|
print(f"β Evaluator failed: {result}") |
|
|
return False |
|
|
else: |
|
|
print(f"β Invalid result format: {result}") |
|
|
return False |
|
|
|
|
|
except Exception as e: |
|
|
print(f"β Evaluator test failed: {e}") |
|
|
return False |
|
|
|
|
|
def test_initial_program(): |
|
|
"""Test the initial program generates parameters""" |
|
|
|
|
|
print("\nπ§ͺ Testing initial program...") |
|
|
|
|
|
try: |
|
|
sys.path.insert(0, '.') |
|
|
from initial_program import optimize_attention |
|
|
|
|
|
params = optimize_attention() |
|
|
|
|
|
print("β
Initial program imported successfully") |
|
|
print(f"π Generated parameters: {list(params.keys())}") |
|
|
|
|
|
|
|
|
required = ['tile_size_m', 'tile_size_n', 'unroll_factor'] |
|
|
for param in required: |
|
|
if param in params: |
|
|
print(f"β
{param}: {params[param]}") |
|
|
else: |
|
|
print(f"β Missing parameter: {param}") |
|
|
return False |
|
|
|
|
|
return True |
|
|
|
|
|
except Exception as e: |
|
|
print(f"β Initial program test failed: {e}") |
|
|
return False |
|
|
|
|
|
def test_mlir_file(): |
|
|
"""Test that the MLIR file exists and is readable""" |
|
|
|
|
|
print("\nπ§ͺ Testing MLIR file...") |
|
|
|
|
|
mlir_file = Path("./mlir/self_attn_with_consts_linalg_dialect.mlir") |
|
|
|
|
|
if mlir_file.exists(): |
|
|
print(f"β
MLIR file exists: {mlir_file}") |
|
|
try: |
|
|
with open(mlir_file, 'r') as f: |
|
|
content = f.read() |
|
|
print(f"β
MLIR file readable: {len(content)} characters") |
|
|
|
|
|
|
|
|
if 'output_shape' in content: |
|
|
print("β
tensor.expand_shape syntax is fixed") |
|
|
else: |
|
|
print("β οΈ tensor.expand_shape may need fixing") |
|
|
|
|
|
return True |
|
|
except Exception as e: |
|
|
print(f"β Cannot read MLIR file: {e}") |
|
|
return False |
|
|
else: |
|
|
print(f"β MLIR file not found: {mlir_file}") |
|
|
return False |
|
|
|
|
|
def main(): |
|
|
"""Run all tests""" |
|
|
|
|
|
print("π Testing Complete AlphaEvolve Setup\n") |
|
|
|
|
|
tests = [ |
|
|
("MLIR File", test_mlir_file), |
|
|
("Initial Program", test_initial_program), |
|
|
("Evaluator", test_evaluator), |
|
|
] |
|
|
|
|
|
results = [] |
|
|
for test_name, test_func in tests: |
|
|
success = test_func() |
|
|
results.append((test_name, success)) |
|
|
|
|
|
|
|
|
print(f"\n{'='*50}") |
|
|
print("TEST SUMMARY") |
|
|
print('='*50) |
|
|
|
|
|
passed = 0 |
|
|
for test_name, success in results: |
|
|
status = "β
PASS" if success else "β FAIL" |
|
|
print(f"{status:8} {test_name}") |
|
|
if success: |
|
|
passed += 1 |
|
|
|
|
|
print(f"\nResults: {passed}/{len(results)} tests passed") |
|
|
|
|
|
if passed == len(results): |
|
|
print("\nπ All tests passed! Ready to run AlphaEvolve!") |
|
|
print("\nπ Run evolution with:") |
|
|
print(" python ../../openevolve-run.py initial_program.py evaluator.py --config config.yaml --iterations 10") |
|
|
print("\nπ― Target: Achieve 32% speedup (1.32x) like AlphaEvolve paper") |
|
|
else: |
|
|
print(f"\nβ οΈ {len(results) - passed} test(s) failed. Fix issues before running evolution.") |
|
|
|
|
|
return passed == len(results) |
|
|
|
|
|
if __name__ == "__main__": |
|
|
success = main() |
|
|
sys.exit(0 if success else 1) |