Spaces:
Running
Running
File size: 7,763 Bytes
7b4f5dd | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 | """
Tests for PerformanceAgent — static scan only (no LLM / GPU required).
"""
from __future__ import annotations
import pathlib
import pytest
from agents.performance_agent import PerformanceAgent
from api.models import OptimizationType
from tools.code_parser import FileEntry
FIXTURES_DIR = pathlib.Path(__file__).parent / "fixtures"
# ──────────────────────────────────────────
# Fixtures
# ──────────────────────────────────────────
@pytest.fixture(scope="module")
def vulnerable_code() -> str:
return (FIXTURES_DIR / "vulnerable_ml_code.py").read_text(encoding="utf-8")
@pytest.fixture(scope="module")
def clean_code() -> str:
return (FIXTURES_DIR / "clean_ml_code.py").read_text(encoding="utf-8")
@pytest.fixture(scope="module")
def agent() -> PerformanceAgent:
return PerformanceAgent()
@pytest.fixture(scope="module")
def vulnerable_files(vulnerable_code: str) -> list[FileEntry]:
return [("vulnerable_ml_code.py", vulnerable_code)]
@pytest.fixture(scope="module")
def perf_findings(agent: PerformanceAgent, vulnerable_files: list[FileEntry]):
return agent.static_scan(vulnerable_files)
# ──────────────────────────────────────────
# Inline test code snippets
# ──────────────────────────────────────────
GPU_LEAK_CODE = '''
import torch
model = load_model().cuda()
def infer(text):
inputs = tokenizer(text, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs)
# Tensor never moved to CPU or deleted — memory leak
return outputs
'''
N_PLUS_ONE_CODE = '''
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("all-MiniLM-L6-v2")
documents = ["doc1", "doc2", "doc3"]
embeddings = []
for doc in documents:
emb = model.encode(doc)
embeddings.append(emb)
'''
FP32_CODE = '''
import torch
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained(
"gpt2",
torch_dtype=torch.float32,
)
'''
NO_GRAD_CODE = '''
import torch
model = load_model()
def predict(text):
inputs = tokenizer(text, return_tensors="pt")
outputs = model(inputs)
return outputs.logits.argmax()
'''
BATCHED_CODE = '''
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("all-MiniLM-L6-v2")
documents = ["doc1", "doc2", "doc3"]
# Correct: batch all at once
embeddings = model.encode(documents, batch_size=32)
'''
# ──────────────────────────────────────────
# Tests
# ──────────────────────────────────────────
class TestGPUMemoryLeakDetection:
def test_detects_gpu_memory_leak(self, agent: PerformanceAgent):
"""Should detect GPU tensor with no corresponding .cpu() or del."""
files: list[FileEntry] = [("test_leak.py", GPU_LEAK_CODE)]
findings = agent.static_scan(files)
gpu_findings = [
f for f in findings
if f.type == OptimizationType.gpu_memory
]
assert len(gpu_findings) > 0, "Expected GPU memory finding for tensor not moved to CPU"
def test_no_leak_with_empty_cache(self, agent: PerformanceAgent):
"""Code that calls empty_cache should produce fewer GPU memory warnings."""
clean_gpu_code = GPU_LEAK_CODE + "\ntorch.cuda.empty_cache()\n"
files: list[FileEntry] = [("clean_gpu.py", clean_gpu_code)]
findings = agent.static_scan(files)
# Should have fewer findings because empty_cache is present
without_cache = agent.static_scan([("test.py", GPU_LEAK_CODE)])
assert len(findings) <= len(without_cache)
class TestNPlusOneEmbeddings:
def test_detects_n_plus_one_embeddings(self, agent: PerformanceAgent):
"""Should detect encode() called inside a for-loop."""
files: list[FileEntry] = [("n_plus_one.py", N_PLUS_ONE_CODE)]
findings = agent.static_scan(files)
throughput_findings = [
f for f in findings
if f.type == OptimizationType.throughput
or "n+1" in f.title.lower()
or "loop" in f.title.lower()
or "batch" in f.suggestion.lower()
]
assert len(throughput_findings) > 0, (
"Expected throughput finding for N+1 embedding calls"
)
def test_no_n_plus_one_for_batch_code(self, agent: PerformanceAgent):
"""Correctly batched embeddings should not be flagged."""
files: list[FileEntry] = [("batched.py", BATCHED_CODE)]
findings = agent.static_scan(files)
n_plus_one_findings = [
f for f in findings
if "n+1" in f.title.lower()
]
assert len(n_plus_one_findings) == 0, "Batched code should not flag N+1"
class TestFP32Inefficiency:
def test_detects_fp32_inefficiency(self, agent: PerformanceAgent):
"""Should detect torch.float32 / .float() usage."""
files: list[FileEntry] = [("fp32_code.py", FP32_CODE)]
findings = agent.static_scan(files)
fp32_findings = [
f for f in findings
if "fp32" in f.title.lower()
or "float32" in f.title.lower()
or "bf16" in f.title.lower()
]
assert len(fp32_findings) > 0, "Expected FP32 inefficiency finding"
def test_fp32_finding_type_is_gpu_memory(self, agent: PerformanceAgent):
files: list[FileEntry] = [("fp32_code.py", FP32_CODE)]
findings = agent.static_scan(files)
fp32_findings = [
f for f in findings
if "fp32" in f.title.lower() or "float32" in f.title.lower()
]
if fp32_findings:
assert fp32_findings[0].type == OptimizationType.gpu_memory
class TestMemorySavingsEstimate:
def test_estimates_memory_savings(self, perf_findings):
"""At least one finding should report a positive savings_mb value."""
savings = [f.saving_mb for f in perf_findings if f.saving_mb and f.saving_mb > 0]
assert len(savings) > 0, (
"Expected at least one finding with savings_mb > 0"
)
def test_total_savings_positive(self, perf_findings):
total = sum(f.saving_mb or 0 for f in perf_findings)
assert total > 0, "Total estimated savings should be > 0 MB"
class TestMissingNoGrad:
def test_detects_missing_no_grad(self, agent: PerformanceAgent):
"""Should detect inference function missing @torch.no_grad."""
files: list[FileEntry] = [("no_grad.py", NO_GRAD_CODE)]
findings = agent.static_scan(files)
no_grad_findings = [
f for f in findings
if "no_grad" in f.title.lower()
or "gradient" in f.suggestion.lower()
]
assert len(no_grad_findings) > 0, "Expected finding for missing @torch.no_grad"
class TestFindingSchema:
def test_all_performance_findings_have_required_fields(self, perf_findings):
for i, finding in enumerate(perf_findings):
assert finding.type is not None, f"Finding {i} missing type"
assert finding.title, f"Finding {i} missing title"
assert finding.suggestion, f"Finding {i} missing suggestion"
def test_vulnerable_code_has_performance_findings(self, perf_findings):
assert len(perf_findings) > 0, (
"PerformanceAgent.static_scan() returned no findings for vulnerable code"
)
|