Spaces:
Sleeping
Sleeping
File size: 32,344 Bytes
dc2106c |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 |
# Copyright (c) ONNX Project Contributors
# SPDX-License-Identifier: Apache-2.0
from __future__ import annotations
import itertools
import os
import pathlib
import tempfile
import unittest
import uuid
from typing import Any
import numpy as np
import parameterized
import onnx
from onnx import ModelProto, TensorProto, checker, helper, shape_inference
from onnx.external_data_helper import (
convert_model_from_external_data,
convert_model_to_external_data,
load_external_data_for_model,
load_external_data_for_tensor,
set_external_data,
)
from onnx.numpy_helper import from_array, to_array
class TestLoadExternalDataBase(unittest.TestCase):
"""Base class for testing external data related behaviors.
Subclasses should be parameterized with a serialization format.
"""
serialization_format: str = "protobuf"
def setUp(self) -> None:
self._temp_dir_obj = tempfile.TemporaryDirectory()
self.temp_dir: str = self._temp_dir_obj.name
self.initializer_value = np.arange(6).reshape(3, 2).astype(np.float32) + 512
self.attribute_value = np.arange(6).reshape(2, 3).astype(np.float32) + 256
self.model_filename = self.create_test_model()
def tearDown(self) -> None:
self._temp_dir_obj.cleanup()
def get_temp_model_filename(self) -> str:
return os.path.join(self.temp_dir, str(uuid.uuid4()) + ".onnx")
def create_external_data_tensor(
self, value: list[Any], tensor_name: str, location: str = ""
) -> TensorProto:
tensor = from_array(np.array(value))
tensor.name = tensor_name
tensor_filename = location or f"{tensor_name}.bin"
set_external_data(tensor, location=tensor_filename)
with open(os.path.join(self.temp_dir, tensor_filename), "wb") as data_file:
data_file.write(tensor.raw_data)
tensor.ClearField("raw_data")
tensor.data_location = onnx.TensorProto.EXTERNAL
return tensor
def create_test_model(self, location: str = "") -> str:
constant_node = onnx.helper.make_node(
"Constant",
inputs=[],
outputs=["values"],
value=self.create_external_data_tensor(
self.attribute_value, "attribute_value" # type: ignore[arg-type]
),
)
initializers = [
self.create_external_data_tensor(
self.initializer_value, "input_value", location # type: ignore[arg-type]
)
]
inputs = [
helper.make_tensor_value_info(
"input_value", onnx.TensorProto.FLOAT, self.initializer_value.shape
)
]
graph = helper.make_graph(
[constant_node],
"test_graph",
inputs=inputs,
outputs=[],
initializer=initializers,
)
model = helper.make_model(graph)
model_filename = os.path.join(self.temp_dir, "model.onnx")
onnx.save_model(model, model_filename, self.serialization_format)
return model_filename
def test_check_model(self) -> None:
if self.serialization_format != "protobuf":
self.skipTest(
"check_model supports protobuf only as binary when provided as a path"
)
checker.check_model(self.model_filename)
@parameterized.parameterized_class(
[
{"serialization_format": "protobuf"},
{"serialization_format": "textproto"},
]
)
class TestLoadExternalData(TestLoadExternalDataBase):
def test_load_external_data(self) -> None:
model = onnx.load_model(self.model_filename, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_load_external_data_for_model(self) -> None:
model = onnx.load_model(
self.model_filename, self.serialization_format, load_external_data=False
)
load_external_data_for_model(model, self.temp_dir)
initializer_tensor = model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_save_external_data(self) -> None:
model = onnx.load_model(self.model_filename, self.serialization_format)
temp_dir = os.path.join(self.temp_dir, "save_copy")
os.mkdir(temp_dir)
new_model_filename = os.path.join(temp_dir, "model.onnx")
onnx.save_model(model, new_model_filename, self.serialization_format)
new_model = onnx.load_model(new_model_filename, self.serialization_format)
initializer_tensor = new_model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = new_model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
@parameterized.parameterized_class(
[
{"serialization_format": "protobuf"},
{"serialization_format": "textproto"},
]
)
class TestLoadExternalDataSingleFile(TestLoadExternalDataBase):
def create_external_data_tensors(
self, tensors_data: list[tuple[list[Any], Any]]
) -> list[TensorProto]:
tensor_filename = "tensors.bin"
tensors = []
with open(os.path.join(self.temp_dir, tensor_filename), "ab") as data_file:
for value, tensor_name in tensors_data:
tensor = from_array(np.array(value))
offset = data_file.tell()
if offset % 4096 != 0:
data_file.write(b"\0" * (4096 - offset % 4096))
offset = offset + 4096 - offset % 4096
data_file.write(tensor.raw_data)
set_external_data(
tensor,
location=tensor_filename,
offset=offset,
length=data_file.tell() - offset,
)
tensor.name = tensor_name
tensor.ClearField("raw_data")
tensor.data_location = onnx.TensorProto.EXTERNAL
tensors.append(tensor)
return tensors
def test_load_external_single_file_data(self) -> None:
model = onnx.load_model(self.model_filename, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_save_external_single_file_data(self) -> None:
model = onnx.load_model(self.model_filename, self.serialization_format)
temp_dir = os.path.join(self.temp_dir, "save_copy")
os.mkdir(temp_dir)
new_model_filename = os.path.join(temp_dir, "model.onnx")
onnx.save_model(model, new_model_filename, self.serialization_format)
new_model = onnx.load_model(new_model_filename, self.serialization_format)
initializer_tensor = new_model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = new_model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
@parameterized.parameterized.expand(itertools.product((True, False), (True, False)))
def test_save_external_invalid_single_file_data_and_check(
self, use_absolute_path: bool, use_model_path: bool
) -> None:
model = onnx.load_model(self.model_filename, self.serialization_format)
model_dir = os.path.join(self.temp_dir, "save_copy")
os.mkdir(model_dir)
traversal_external_data_dir = os.path.join(
self.temp_dir, "invlid_external_data"
)
os.mkdir(traversal_external_data_dir)
if use_absolute_path:
traversal_external_data_location = os.path.join(
traversal_external_data_dir, "tensors.bin"
)
else:
traversal_external_data_location = "../invlid_external_data/tensors.bin"
external_data_dir = os.path.join(self.temp_dir, "external_data")
os.mkdir(external_data_dir)
new_model_filepath = os.path.join(model_dir, "model.onnx")
def convert_model_to_external_data_no_check(model: ModelProto, location: str):
for tensor in model.graph.initializer:
if tensor.HasField("raw_data"):
set_external_data(tensor, location)
convert_model_to_external_data_no_check(
model,
location=traversal_external_data_location,
)
onnx.save_model(model, new_model_filepath, self.serialization_format)
if use_model_path:
with self.assertRaises(onnx.checker.ValidationError):
_ = onnx.load_model(new_model_filepath, self.serialization_format)
else:
onnx_model = onnx.load_model(
new_model_filepath, self.serialization_format, load_external_data=False
)
with self.assertRaises(onnx.checker.ValidationError):
load_external_data_for_model(onnx_model, external_data_dir)
@parameterized.parameterized_class(
[
{"serialization_format": "protobuf"},
{"serialization_format": "textproto"},
]
)
class TestSaveAllTensorsAsExternalData(unittest.TestCase):
serialization_format: str = "protobuf"
def setUp(self) -> None:
self._temp_dir_obj = tempfile.TemporaryDirectory()
self.temp_dir: str = self._temp_dir_obj.name
self.initializer_value = np.arange(6).reshape(3, 2).astype(np.float32) + 512
self.attribute_value = np.arange(6).reshape(2, 3).astype(np.float32) + 256
self.model = self.create_test_model_proto()
def get_temp_model_filename(self):
return os.path.join(self.temp_dir, str(uuid.uuid4()) + ".onnx")
def create_data_tensors(
self, tensors_data: list[tuple[list[Any], Any]]
) -> list[TensorProto]:
tensors = []
for value, tensor_name in tensors_data:
tensor = from_array(np.array(value))
tensor.name = tensor_name
tensors.append(tensor)
return tensors
def create_test_model_proto(self) -> ModelProto:
tensors = self.create_data_tensors(
[
(self.attribute_value, "attribute_value"), # type: ignore[list-item]
(self.initializer_value, "input_value"), # type: ignore[list-item]
]
)
constant_node = onnx.helper.make_node(
"Constant", inputs=[], outputs=["values"], value=tensors[0]
)
inputs = [
helper.make_tensor_value_info(
"input_value", onnx.TensorProto.FLOAT, self.initializer_value.shape
)
]
graph = helper.make_graph(
[constant_node],
"test_graph",
inputs=inputs,
outputs=[],
initializer=[tensors[1]],
)
return helper.make_model(graph)
@unittest.skipIf(
serialization_format != "protobuf",
"check_model supports protobuf only when provided as a path",
)
def test_check_model(self) -> None:
checker.check_model(self.model)
def test_convert_model_to_external_data_with_size_threshold(self) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(self.model, size_threshold=1024)
onnx.save_model(self.model, model_file_path, self.serialization_format)
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertFalse(initializer_tensor.HasField("data_location"))
def test_convert_model_to_external_data_without_size_threshold(self) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(self.model, size_threshold=0)
onnx.save_model(self.model, model_file_path, self.serialization_format)
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertTrue(initializer_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
def test_convert_model_to_external_data_from_one_file_with_location(self) -> None:
model_file_path = self.get_temp_model_filename()
external_data_file = str(uuid.uuid4())
convert_model_to_external_data(
self.model,
size_threshold=0,
all_tensors_to_one_file=True,
location=external_data_file,
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, external_data_file)))
model = onnx.load_model(model_file_path, self.serialization_format)
# test convert model from external data
convert_model_from_external_data(model)
model_file_path = self.get_temp_model_filename()
onnx.save_model(model, model_file_path, self.serialization_format)
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertFalse(len(initializer_tensor.external_data))
self.assertEqual(initializer_tensor.data_location, TensorProto.DEFAULT)
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
self.assertFalse(len(attribute_tensor.external_data))
self.assertEqual(attribute_tensor.data_location, TensorProto.DEFAULT)
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_convert_model_to_external_data_from_one_file_without_location_uses_model_name(
self,
) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(
self.model, size_threshold=0, all_tensors_to_one_file=True
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
self.assertTrue(os.path.isfile(model_file_path))
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, model_file_path)))
def test_convert_model_to_external_data_one_file_per_tensor_without_attribute(
self,
) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(
self.model,
size_threshold=0,
all_tensors_to_one_file=False,
convert_attribute=False,
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
self.assertTrue(os.path.isfile(model_file_path))
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, "input_value")))
self.assertFalse(os.path.isfile(os.path.join(self.temp_dir, "attribute_value")))
def test_convert_model_to_external_data_one_file_per_tensor_with_attribute(
self,
) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(
self.model,
size_threshold=0,
all_tensors_to_one_file=False,
convert_attribute=True,
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
self.assertTrue(os.path.isfile(model_file_path))
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, "input_value")))
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, "attribute_value")))
def test_convert_model_to_external_data_does_not_convert_attribute_values(
self,
) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(
self.model,
size_threshold=0,
convert_attribute=False,
all_tensors_to_one_file=False,
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
self.assertTrue(os.path.isfile(os.path.join(self.temp_dir, "input_value")))
self.assertFalse(os.path.isfile(os.path.join(self.temp_dir, "attribute_value")))
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertTrue(initializer_tensor.HasField("data_location"))
attribute_tensor = model.graph.node[0].attribute[0].t
self.assertFalse(attribute_tensor.HasField("data_location"))
def test_convert_model_to_external_data_converts_attribute_values(self) -> None:
model_file_path = self.get_temp_model_filename()
convert_model_to_external_data(
self.model, size_threshold=0, convert_attribute=True
)
onnx.save_model(self.model, model_file_path, self.serialization_format)
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
self.assertTrue(initializer_tensor.HasField("data_location"))
attribute_tensor = model.graph.node[0].attribute[0].t
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
self.assertTrue(attribute_tensor.HasField("data_location"))
def test_save_model_does_not_convert_to_external_data_and_saves_the_model(
self,
) -> None:
model_file_path = self.get_temp_model_filename()
onnx.save_model(
self.model,
model_file_path,
self.serialization_format,
save_as_external_data=False,
)
self.assertTrue(os.path.isfile(model_file_path))
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertFalse(initializer_tensor.HasField("data_location"))
attribute_tensor = model.graph.node[0].attribute[0].t
self.assertFalse(attribute_tensor.HasField("data_location"))
def test_save_model_does_convert_and_saves_the_model(self) -> None:
model_file_path = self.get_temp_model_filename()
onnx.save_model(
self.model,
model_file_path,
self.serialization_format,
save_as_external_data=True,
all_tensors_to_one_file=True,
location=None,
size_threshold=0,
convert_attribute=False,
)
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertTrue(initializer_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
self.assertFalse(attribute_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_save_model_without_loading_external_data(self) -> None:
model_file_path = self.get_temp_model_filename()
onnx.save_model(
self.model,
model_file_path,
self.serialization_format,
save_as_external_data=True,
location=None,
size_threshold=0,
convert_attribute=False,
)
# Save without load_external_data
model = onnx.load_model(
model_file_path, self.serialization_format, load_external_data=False
)
onnx.save_model(
model,
model_file_path,
self.serialization_format,
save_as_external_data=True,
location=None,
size_threshold=0,
convert_attribute=False,
)
# Load the saved model again; Only works if the saved path is under the same directory
model = onnx.load_model(model_file_path, self.serialization_format)
initializer_tensor = model.graph.initializer[0]
self.assertTrue(initializer_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(initializer_tensor), self.initializer_value)
attribute_tensor = model.graph.node[0].attribute[0].t
self.assertFalse(attribute_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(attribute_tensor), self.attribute_value)
def test_save_model_with_existing_raw_data_should_override(self) -> None:
model_file_path = self.get_temp_model_filename()
original_raw_data = self.model.graph.initializer[0].raw_data
onnx.save_model(
self.model,
model_file_path,
self.serialization_format,
save_as_external_data=True,
size_threshold=0,
)
self.assertTrue(os.path.isfile(model_file_path))
model = onnx.load_model(
model_file_path, self.serialization_format, load_external_data=False
)
initializer_tensor = model.graph.initializer[0]
initializer_tensor.raw_data = b"dummpy_raw_data"
# If raw_data and external tensor exist at the same time, override existing raw_data
load_external_data_for_tensor(initializer_tensor, self.temp_dir)
self.assertEqual(initializer_tensor.raw_data, original_raw_data)
@parameterized.parameterized_class(
[
{"serialization_format": "protobuf"},
{"serialization_format": "textproto"},
]
)
class TestExternalDataToArray(unittest.TestCase):
serialization_format: str = "protobuf"
def setUp(self) -> None:
self._temp_dir_obj = tempfile.TemporaryDirectory()
self.temp_dir: str = self._temp_dir_obj.name
self._model_file_path: str = os.path.join(self.temp_dir, "model.onnx")
self.large_data = np.random.rand(10, 60, 100).astype(np.float32)
self.small_data = (200, 300)
self.model = self.create_test_model()
@property
def model_file_path(self):
return self._model_file_path
def tearDown(self) -> None:
self._temp_dir_obj.cleanup()
def create_test_model(self) -> ModelProto:
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, self.large_data.shape)
input_init = helper.make_tensor(
name="X",
data_type=TensorProto.FLOAT,
dims=self.large_data.shape,
vals=self.large_data.tobytes(),
raw=True,
)
shape_data = np.array(self.small_data, np.int64)
shape_init = helper.make_tensor(
name="Shape",
data_type=TensorProto.INT64,
dims=shape_data.shape,
vals=shape_data.tobytes(),
raw=True,
)
C = helper.make_tensor_value_info("C", TensorProto.INT64, self.small_data)
reshape = onnx.helper.make_node(
"Reshape",
inputs=["X", "Shape"],
outputs=["Y"],
)
cast = onnx.helper.make_node(
"Cast", inputs=["Y"], outputs=["C"], to=TensorProto.INT64
)
graph_def = helper.make_graph(
[reshape, cast],
"test-model",
[X],
[C],
initializer=[input_init, shape_init],
)
model = helper.make_model(graph_def, producer_name="onnx-example")
return model
@unittest.skipIf(
serialization_format != "protobuf",
"check_model supports protobuf only when provided as a path",
)
def test_check_model(self) -> None:
checker.check_model(self.model)
def test_reshape_inference_with_external_data_fail(self) -> None:
onnx.save_model(
self.model,
self.model_file_path,
self.serialization_format,
save_as_external_data=True,
all_tensors_to_one_file=False,
size_threshold=0,
)
model_without_external_data = onnx.load(
self.model_file_path, self.serialization_format, load_external_data=False
)
# Shape inference of Reshape uses ParseData
# ParseData cannot handle external data and should throw the error as follows:
# Cannot parse data from external tensors. Please load external data into raw data for tensor: Shape
self.assertRaises(
shape_inference.InferenceError,
shape_inference.infer_shapes,
model_without_external_data,
strict_mode=True,
)
def test_to_array_with_external_data(self) -> None:
onnx.save_model(
self.model,
self.model_file_path,
self.serialization_format,
save_as_external_data=True,
all_tensors_to_one_file=False,
size_threshold=0,
)
# raw_data of external tensor is not loaded
model = onnx.load(
self.model_file_path, self.serialization_format, load_external_data=False
)
# Specify self.temp_dir to load external tensor
loaded_large_data = to_array(model.graph.initializer[0], self.temp_dir)
np.testing.assert_allclose(loaded_large_data, self.large_data)
def test_save_model_with_external_data_multiple_times(self) -> None:
# Test onnx.save should respectively handle typical tensor and external tensor properly
# 1st save: save two tensors which have raw_data
# Only w_large will be stored as external tensors since it's larger than 1024
onnx.save_model(
self.model,
self.model_file_path,
self.serialization_format,
save_as_external_data=True,
all_tensors_to_one_file=False,
location=None,
size_threshold=1024,
convert_attribute=True,
)
model_without_loading_external = onnx.load(
self.model_file_path, self.serialization_format, load_external_data=False
)
large_input_tensor = model_without_loading_external.graph.initializer[0]
self.assertTrue(large_input_tensor.HasField("data_location"))
np.testing.assert_allclose(
to_array(large_input_tensor, self.temp_dir), self.large_data
)
small_shape_tensor = model_without_loading_external.graph.initializer[1]
self.assertTrue(not small_shape_tensor.HasField("data_location"))
np.testing.assert_allclose(to_array(small_shape_tensor), self.small_data)
# 2nd save: one tensor has raw_data (small); one external tensor (large)
# Save them both as external tensors this time
onnx.save_model(
model_without_loading_external,
self.model_file_path,
self.serialization_format,
save_as_external_data=True,
all_tensors_to_one_file=False,
location=None,
size_threshold=0,
convert_attribute=True,
)
model_without_loading_external = onnx.load(
self.model_file_path, self.serialization_format, load_external_data=False
)
large_input_tensor = model_without_loading_external.graph.initializer[0]
self.assertTrue(large_input_tensor.HasField("data_location"))
np.testing.assert_allclose(
to_array(large_input_tensor, self.temp_dir), self.large_data
)
small_shape_tensor = model_without_loading_external.graph.initializer[1]
self.assertTrue(small_shape_tensor.HasField("data_location"))
np.testing.assert_allclose(
to_array(small_shape_tensor, self.temp_dir), self.small_data
)
class TestNotAllowToLoadExternalDataOutsideModelDirectory(TestLoadExternalDataBase):
"""Essential test to check that onnx (validate) C++ code will not allow to load external_data outside the model
directory.
"""
def create_external_data_tensor(
self, value: list[Any], tensor_name: str, location: str = ""
) -> TensorProto:
tensor = from_array(np.array(value))
tensor.name = tensor_name
tensor_filename = location or f"{tensor_name}.bin"
set_external_data(tensor, location=tensor_filename)
tensor.ClearField("raw_data")
tensor.data_location = onnx.TensorProto.EXTERNAL
return tensor
def test_check_model(self) -> None:
"""We only test the model validation as onnxruntime uses this to load the model."""
self.model_filename = self.create_test_model("../../file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
def test_check_model_relative(self) -> None:
"""More relative path test."""
self.model_filename = self.create_test_model("../test/../file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
def test_check_model_absolute(self) -> None:
"""ONNX checker disallows using absolute path as location in external tensor."""
self.model_filename = self.create_test_model("//file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
@unittest.skipIf(os.name != "nt", reason="Skip Windows test")
class TestNotAllowToLoadExternalDataOutsideModelDirectoryOnWindows(
TestNotAllowToLoadExternalDataOutsideModelDirectory
):
"""Essential test to check that onnx (validate) C++ code will not allow to load external_data outside the model
directory.
"""
def test_check_model(self) -> None:
"""We only test the model validation as onnxruntime uses this to load the model."""
self.model_filename = self.create_test_model("..\\..\\file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
def test_check_model_relative(self) -> None:
"""More relative path test."""
self.model_filename = self.create_test_model("..\\test\\..\\file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
def test_check_model_absolute(self) -> None:
"""ONNX checker disallows using absolute path as location in external tensor."""
self.model_filename = self.create_test_model("C:/file.bin")
with self.assertRaises(onnx.checker.ValidationError):
checker.check_model(self.model_filename)
class TestSaveAllTensorsAsExternalDataWithPath(TestSaveAllTensorsAsExternalData):
def get_temp_model_filename(self) -> pathlib.Path:
return pathlib.Path(super().get_temp_model_filename())
class TestExternalDataToArrayWithPath(TestExternalDataToArray):
@property
def model_file_path(self) -> pathlib.Path:
return pathlib.Path(self._model_file_path)
if __name__ == "__main__":
unittest.main()
|