EngineX-Cambricon/enginex-mlu370-any2any
9
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init

Browse Source
This commit is contained in:
luopingyi
2025-10-09 16:47:16 +08:00
parent c8feb4deb5
commit e27e3f16bb
5248 changed files with 1778505 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

0
transformers/tests/models/zoedepth/__init__.py Normal file
View File

251
transformers/tests/models/zoedepth/test_image_processing_zoedepth.py Normal file
View File

@@ -0,0 +1,251 @@
# Copyright 2024 HuggingFace Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
from dataclasses import dataclass
import numpy as np
from transformers.testing_utils import require_torch, require_vision
from transformers.utils import is_torch_available, is_torchvision_available, is_vision_available
from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
if is_torch_available():
import torch
if is_vision_available():
from transformers import ZoeDepthImageProcessor
if is_torchvision_available():
from transformers import ZoeDepthImageProcessorFast
@dataclass
class ZoeDepthDepthOutputProxy:
predicted_depth: torch.FloatTensor = None
class ZoeDepthImageProcessingTester:
def __init__(
self,
parent,
batch_size=7,
num_channels=3,
image_size=18,
min_resolution=30,
max_resolution=400,
do_resize=True,
size=None,
ensure_multiple_of=32,
keep_aspect_ratio=False,
do_normalize=True,
image_mean=[0.5, 0.5, 0.5],
image_std=[0.5, 0.5, 0.5],
do_pad=True,
):
size = size if size is not None else {"height": 18, "width": 18}
self.parent = parent
self.batch_size = batch_size
self.num_channels = num_channels
self.image_size = image_size
self.min_resolution = min_resolution
self.max_resolution = max_resolution
self.do_resize = do_resize
self.size = size
self.ensure_multiple_of = ensure_multiple_of
self.keep_aspect_ratio = keep_aspect_ratio
self.do_normalize = do_normalize
self.image_mean = image_mean
self.image_std = image_std
self.do_pad = do_pad
def prepare_image_processor_dict(self):
return {
"do_resize": self.do_resize,
"size": self.size,
"ensure_multiple_of": self.ensure_multiple_of,
"keep_aspect_ratio": self.keep_aspect_ratio,
"do_normalize": self.do_normalize,
"image_mean": self.image_mean,
"image_std": self.image_std,
"do_pad": self.do_pad,
}
def expected_output_image_shape(self, images):
return self.num_channels, self.ensure_multiple_of, self.ensure_multiple_of
def prepare_image_inputs(self, equal_resolution=False, numpify=False, torchify=False):
return prepare_image_inputs(
batch_size=self.batch_size,
num_channels=self.num_channels,
min_resolution=self.min_resolution,
max_resolution=self.max_resolution,
equal_resolution=equal_resolution,
numpify=numpify,
torchify=torchify,
)
def prepare_depth_outputs(self):
depth_tensors = prepare_image_inputs(
batch_size=self.batch_size,
num_channels=1,
min_resolution=self.min_resolution,
max_resolution=self.max_resolution,
equal_resolution=True,
torchify=True,
)
depth_tensors = [depth_tensor.squeeze(0) for depth_tensor in depth_tensors]
stacked_depth_tensors = torch.stack(depth_tensors, dim=0)
return ZoeDepthDepthOutputProxy(predicted_depth=stacked_depth_tensors)
@require_torch
@require_vision
class ZoeDepthImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
image_processing_class = ZoeDepthImageProcessor if is_vision_available() else None
fast_image_processing_class = ZoeDepthImageProcessorFast if is_torchvision_available() else None
def setUp(self):
super().setUp()
self.image_processor_tester = ZoeDepthImageProcessingTester(self)
@property
def image_processor_dict(self):
return self.image_processor_tester.prepare_image_processor_dict()
def test_image_processor_properties(self):
image_processing = self.image_processing_class(**self.image_processor_dict)
self.assertTrue(hasattr(image_processing, "image_mean"))
self.assertTrue(hasattr(image_processing, "image_std"))
self.assertTrue(hasattr(image_processing, "do_normalize"))
self.assertTrue(hasattr(image_processing, "do_resize"))
self.assertTrue(hasattr(image_processing, "size"))
self.assertTrue(hasattr(image_processing, "ensure_multiple_of"))
self.assertTrue(hasattr(image_processing, "do_rescale"))
self.assertTrue(hasattr(image_processing, "rescale_factor"))
self.assertTrue(hasattr(image_processing, "do_pad"))
def test_image_processor_from_dict_with_kwargs(self):
for image_processing_class in self.image_processor_list:
image_processor = image_processing_class(**self.image_processor_dict)
self.assertEqual(image_processor.size, {"height": 18, "width": 18})
for image_processing_class in self.image_processor_list:
modified_dict = self.image_processor_dict
modified_dict["size"] = 42
image_processor = image_processing_class(**modified_dict)
self.assertEqual(image_processor.size, {"height": 42, "width": 42})
def test_ensure_multiple_of(self):
# Test variable by turning off all other variables which affect the size, size which is not multiple of 32
image = np.zeros((489, 640, 3))
size = {"height": 380, "width": 513}
multiple = 32
for image_processor_class in self.image_processor_list:
image_processor = image_processor_class(
do_pad=False, ensure_multiple_of=multiple, size=size, keep_aspect_ratio=False
)
pixel_values = image_processor(image, return_tensors="pt").pixel_values
self.assertEqual(list(pixel_values.shape), [1, 3, 384, 512])
self.assertTrue(pixel_values.shape[2] % multiple == 0)
self.assertTrue(pixel_values.shape[3] % multiple == 0)
# Test variable by turning off all other variables which affect the size, size which is already multiple of 32
image = np.zeros((511, 511, 3))
height, width = 512, 512
size = {"height": height, "width": width}
multiple = 32
for image_processor_class in self.image_processor_list:
image_processor = image_processor_class(
do_pad=False, ensure_multiple_of=multiple, size=size, keep_aspect_ratio=False
)
pixel_values = image_processor(image, return_tensors="pt").pixel_values
self.assertEqual(list(pixel_values.shape), [1, 3, height, width])
self.assertTrue(pixel_values.shape[2] % multiple == 0)
self.assertTrue(pixel_values.shape[3] % multiple == 0)
def test_keep_aspect_ratio(self):
# Test `keep_aspect_ratio=True` by turning off all other variables which affect the size
height, width = 489, 640
image = np.zeros((height, width, 3))
size = {"height": 512, "width": 512}
for image_processor_class in self.image_processor_list:
image_processor = image_processor_class(
do_pad=False, keep_aspect_ratio=True, size=size, ensure_multiple_of=1
)
pixel_values = image_processor(image, return_tensors="pt").pixel_values
# As can be seen, the image is resized to the maximum size that fits in the specified size
self.assertEqual(list(pixel_values.shape), [1, 3, 512, 670])
# Test `keep_aspect_ratio=False` by turning off all other variables which affect the size
for image_processor_class in self.image_processor_list:
image_processor = image_processor_class(
do_pad=False, keep_aspect_ratio=False, size=size, ensure_multiple_of=1
)
pixel_values = image_processor(image, return_tensors="pt").pixel_values
# As can be seen, the size is respected
self.assertEqual(list(pixel_values.shape), [1, 3, size["height"], size["width"]])
# Test `keep_aspect_ratio=True` with `ensure_multiple_of` set
image = np.zeros((489, 640, 3))
size = {"height": 511, "width": 511}
multiple = 32
for image_processor_class in self.image_processor_list:
image_processor = image_processor_class(size=size, keep_aspect_ratio=True, ensure_multiple_of=multiple)
pixel_values = image_processor(image, return_tensors="pt").pixel_values
self.assertEqual(list(pixel_values.shape), [1, 3, 512, 672])
self.assertTrue(pixel_values.shape[2] % multiple == 0)
self.assertTrue(pixel_values.shape[3] % multiple == 0)
# extend this test to check if removal of padding works fine!
def test_post_processing_equivalence(self):
outputs = self.image_processor_tester.prepare_depth_outputs()
image_processor_fast = self.fast_image_processing_class(**self.image_processor_dict)
image_processor_slow = self.image_processing_class(**self.image_processor_dict)
source_sizes = [outputs.predicted_depth.shape[1:]] * self.image_processor_tester.batch_size
target_sizes = [
torch.Size([outputs.predicted_depth.shape[1] // 2, *(outputs.predicted_depth.shape[2:])])
] * self.image_processor_tester.batch_size
processed_fast = image_processor_fast.post_process_depth_estimation(
outputs,
source_sizes=source_sizes,
target_sizes=target_sizes,
)
processed_slow = image_processor_slow.post_process_depth_estimation(
outputs,
source_sizes=source_sizes,
target_sizes=target_sizes,
)
for pred_fast, pred_slow in zip(processed_fast, processed_slow):
depth_fast = pred_fast["predicted_depth"]
depth_slow = pred_slow["predicted_depth"]
torch.testing.assert_close(depth_fast, depth_slow, atol=1e-1, rtol=1e-3)
self.assertLessEqual(torch.mean(torch.abs(depth_fast.float() - depth_slow.float())).item(), 5e-3)

350
transformers/tests/models/zoedepth/test_modeling_zoedepth.py Normal file
View File

@@ -0,0 +1,350 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Testing suite for the PyTorch ZoeDepth model."""
import unittest
import numpy as np
from transformers import Dinov2Config, ZoeDepthConfig
from transformers.file_utils import is_torch_available, is_vision_available
from transformers.testing_utils import require_torch, require_vision, slow, torch_device
from transformers.utils.import_utils import get_torch_major_and_minor_version
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import ZoeDepthForDepthEstimation
if is_vision_available():
from PIL import Image
from transformers import ZoeDepthImageProcessor
class ZoeDepthModelTester:
def __init__(
self,
parent,
batch_size=2,
num_channels=3,
image_size=32,
patch_size=16,
use_labels=True,
num_labels=3,
is_training=True,
hidden_size=4,
num_hidden_layers=2,
num_attention_heads=2,
intermediate_size=8,
out_features=["stage1", "stage2"],
apply_layernorm=False,
reshape_hidden_states=False,
neck_hidden_sizes=[2, 2],
fusion_hidden_size=6,
bottleneck_features=6,
num_out_features=[6, 6, 6, 6],
):
self.parent = parent
self.batch_size = batch_size
self.num_channels = num_channels
self.image_size = image_size
self.patch_size = patch_size
self.hidden_size = hidden_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.intermediate_size = intermediate_size
self.out_features = out_features
self.apply_layernorm = apply_layernorm
self.reshape_hidden_states = reshape_hidden_states
self.use_labels = use_labels
self.num_labels = num_labels
self.is_training = is_training
self.neck_hidden_sizes = neck_hidden_sizes
self.fusion_hidden_size = fusion_hidden_size
self.bottleneck_features = bottleneck_features
self.num_out_features = num_out_features
# ZoeDepth's sequence length
self.seq_length = (self.image_size // self.patch_size) ** 2 + 1
def prepare_config_and_inputs(self):
pixel_values = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size])
labels = None
if self.use_labels:
labels = ids_tensor([self.batch_size, self.image_size, self.image_size], self.num_labels)
config = self.get_config()
return config, pixel_values, labels
def get_config(self):
return ZoeDepthConfig(
backbone_config=self.get_backbone_config(),
backbone=None,
neck_hidden_sizes=self.neck_hidden_sizes,
fusion_hidden_size=self.fusion_hidden_size,
bottleneck_features=self.bottleneck_features,
num_out_features=self.num_out_features,
)
def get_backbone_config(self):
return Dinov2Config(
image_size=self.image_size,
patch_size=self.patch_size,
num_channels=self.num_channels,
hidden_size=self.hidden_size,
num_hidden_layers=self.num_hidden_layers,
num_attention_heads=self.num_attention_heads,
intermediate_size=self.intermediate_size,
is_training=self.is_training,
out_features=self.out_features,
reshape_hidden_states=self.reshape_hidden_states,
)
def create_and_check_for_depth_estimation(self, config, pixel_values, labels):
config.num_labels = self.num_labels
model = ZoeDepthForDepthEstimation(config)
model.to(torch_device)
model.eval()
result = model(pixel_values)
self.parent.assertEqual(result.predicted_depth.shape, (self.batch_size, self.image_size, self.image_size))
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
config, pixel_values, labels = config_and_inputs
inputs_dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class ZoeDepthModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
"""
Here we also overwrite some of the tests of test_modeling_common.py, as ZoeDepth does not use input_ids, inputs_embeds,
attention_mask and seq_length.
"""
all_model_classes = (ZoeDepthForDepthEstimation,) if is_torch_available() else ()
pipeline_model_mapping = {"depth-estimation": ZoeDepthForDepthEstimation} if is_torch_available() else {}
test_pruning = False
test_resize_embeddings = False
test_head_masking = False
# `strict=True/False` are both failing with torch 2.7, see #38677
test_torch_exportable = get_torch_major_and_minor_version() != "2.7"
def setUp(self):
self.model_tester = ZoeDepthModelTester(self)
self.config_tester = ConfigTester(
self, config_class=ZoeDepthConfig, has_text_modality=False, hidden_size=37, common_properties=[]
)
def test_config(self):
self.config_tester.run_common_tests()
@unittest.skip(reason="ZoeDepth with AutoBackbone does not have a base model and hence no input_embeddings")
def test_inputs_embeds(self):
pass
@unittest.skip(reason="ZoeDepth with AutoBackbone does not have a base model and hence no input_embeddings")
def test_model_get_set_embeddings(self):
pass
def test_for_depth_estimation(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_depth_estimation(*config_and_inputs)
@unittest.skip(reason="ZoeDepth with AutoBackbone does not have a base model and hence no input_embeddings")
def test_model_common_attributes(self):
pass
@unittest.skip(reason="ZoeDepth does not support training yet")
def test_training(self):
pass
@unittest.skip(reason="ZoeDepth does not support training yet")
def test_training_gradient_checkpointing(self):
pass
@unittest.skip(reason="ZoeDepth does not support training yet")
def test_training_gradient_checkpointing_use_reentrant(self):
pass
@unittest.skip(reason="ZoeDepth does not support training yet")
def test_training_gradient_checkpointing_use_reentrant_false(self):
pass
@slow
def test_model_from_pretrained(self):
model_name = "Intel/zoedepth-nyu"
model = ZoeDepthForDepthEstimation.from_pretrained(model_name)
self.assertIsNotNone(model)
# We will verify our results on an image of cute cats
def prepare_img():
image = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png")
return image
@require_torch
@require_vision
@slow
class ZoeDepthModelIntegrationTest(unittest.TestCase):
expected_slice_post_processing = {
(False, False): [
[[1.1348238, 1.1193453, 1.130562], [1.1754476, 1.1613507, 1.1701596], [1.2287744, 1.2101802, 1.2148322]],
[[2.7170, 2.6550, 2.6839], [2.9827, 2.9438, 2.9587], [3.2340, 3.1817, 3.1602]],
],
(False, True): [
[[1.0610938, 1.1042216, 1.1429265], [1.1099341, 1.148696, 1.1817775], [1.1656011, 1.1988826, 1.2268101]],
[[2.5848, 2.7391, 2.8694], [2.7882, 2.9872, 3.1244], [2.9436, 3.1812, 3.3188]],
],
(True, False): [
[[1.8382794, 1.8380532, 1.8375976], [1.848761, 1.8485023, 1.8479986], [1.8571457, 1.8568444, 1.8562847]],
[[6.2030, 6.1902, 6.1777], [6.2303, 6.2176, 6.2053], [6.2561, 6.2436, 6.2312]],
],
(True, True): [
[[1.8306141, 1.8305621, 1.8303483], [1.8410318, 1.8409299, 1.8406585], [1.8492792, 1.8491366, 1.8488203]],
[[6.2616, 6.2520, 6.2435], [6.2845, 6.2751, 6.2667], [6.3065, 6.2972, 6.2887]],
],
} # (pad, flip)
def test_inference_depth_estimation(self):
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu")
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu").to(torch_device)
image = prepare_img()
inputs = image_processor(images=image, return_tensors="pt").to(torch_device)
# forward pass
with torch.no_grad():
outputs = model(**inputs)
predicted_depth = outputs.predicted_depth
# verify the predicted depth
expected_shape = torch.Size((1, 384, 512))
self.assertEqual(predicted_depth.shape, expected_shape)
expected_slice = torch.tensor(
[[1.0020, 1.0219, 1.0389], [1.0349, 1.0816, 1.1000], [1.0576, 1.1094, 1.1249]],
).to(torch_device)
torch.testing.assert_close(outputs.predicted_depth[0, :3, :3], expected_slice, rtol=1e-4, atol=1e-4)
def test_inference_depth_estimation_multiple_heads(self):
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu-kitti")
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
image = prepare_img()
inputs = image_processor(images=image, return_tensors="pt").to(torch_device)
# forward pass
with torch.no_grad():
outputs = model(**inputs)
predicted_depth = outputs.predicted_depth
# verify the predicted depth
expected_shape = torch.Size((1, 384, 512))
self.assertEqual(predicted_depth.shape, expected_shape)
expected_slice = torch.tensor(
[[1.1571, 1.1438, 1.1783], [1.2163, 1.2036, 1.2320], [1.2688, 1.2461, 1.2734]],
).to(torch_device)
torch.testing.assert_close(outputs.predicted_depth[0, :3, :3], expected_slice, rtol=1e-4, atol=1e-4)
def check_target_size(
self,
image_processor,
pad_input,
images,
outputs,
raw_outputs,
raw_outputs_flipped=None,
):
outputs_large = image_processor.post_process_depth_estimation(
raw_outputs,
[img.size[::-1] for img in images],
outputs_flipped=raw_outputs_flipped,
target_sizes=[tuple(np.array(img.size[::-1]) * 2) for img in images],
do_remove_padding=pad_input,
)
for img, out, out_l in zip(images, outputs, outputs_large):
out = out["predicted_depth"]
out_l = out_l["predicted_depth"]
out_l_reduced = torch.nn.functional.interpolate(
out_l.unsqueeze(0).unsqueeze(1), size=img.size[::-1], mode="bicubic", align_corners=False
)
out_l_reduced = out_l_reduced.squeeze(0).squeeze(0)
torch.testing.assert_close(out, out_l_reduced, rtol=2e-2, atol=2e-2)
def check_post_processing_test(self, image_processor, images, model, pad_input=True, flip_aug=True):
inputs = image_processor(images=images, return_tensors="pt", do_pad=pad_input).to(torch_device)
with torch.no_grad():
raw_outputs = model(**inputs)
raw_outputs_flipped = None
if flip_aug:
raw_outputs_flipped = model(pixel_values=torch.flip(inputs.pixel_values, dims=[3]))
outputs = image_processor.post_process_depth_estimation(
raw_outputs,
[img.size[::-1] for img in images],
outputs_flipped=raw_outputs_flipped,
do_remove_padding=pad_input,
)
expected_slices = torch.tensor(self.expected_slice_post_processing[pad_input, flip_aug]).to(torch_device)
for img, out, expected_slice in zip(images, outputs, expected_slices):
out = out["predicted_depth"]
self.assertTrue(img.size == out.shape[::-1])
torch.testing.assert_close(expected_slice, out[:3, :3], rtol=1e-3, atol=1e-3)
self.check_target_size(image_processor, pad_input, images, outputs, raw_outputs, raw_outputs_flipped)
def test_post_processing_depth_estimation_post_processing_nopad_noflip(self):
images = [prepare_img(), Image.open("./tests/fixtures/tests_samples/COCO/000000004016.png")]
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu-kitti", keep_aspect_ratio=False)
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
self.check_post_processing_test(image_processor, images, model, pad_input=False, flip_aug=False)
def test_inference_depth_estimation_post_processing_nopad_flip(self):
images = [prepare_img(), Image.open("./tests/fixtures/tests_samples/COCO/000000004016.png")]
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu-kitti", keep_aspect_ratio=False)
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
self.check_post_processing_test(image_processor, images, model, pad_input=False, flip_aug=True)
def test_inference_depth_estimation_post_processing_pad_noflip(self):
images = [prepare_img(), Image.open("./tests/fixtures/tests_samples/COCO/000000004016.png")]
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu-kitti", keep_aspect_ratio=False)
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
self.check_post_processing_test(image_processor, images, model, pad_input=True, flip_aug=False)
def test_inference_depth_estimation_post_processing_pad_flip(self):
images = [prepare_img(), Image.open("./tests/fixtures/tests_samples/COCO/000000004016.png")]
image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu-kitti", keep_aspect_ratio=False)
model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
self.check_post_processing_test(image_processor, images, model, pad_input=True, flip_aug=True)