Spaces:
Paused
Paused
File size: 14,468 Bytes
ee6e328 |
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 |
# coding=utf-8
# Copyright 2022 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 GLPN model. """
import inspect
import unittest
from transformers import is_torch_available, is_vision_available
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_vision, slow, torch_device
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 MODEL_MAPPING, GLPNConfig, GLPNForDepthEstimation, GLPNModel
from transformers.models.glpn.modeling_glpn import GLPN_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
from PIL import Image
from transformers import GLPNImageProcessor
class GLPNConfigTester(ConfigTester):
def create_and_test_config_common_properties(self):
config = self.config_class(**self.inputs_dict)
self.parent.assertTrue(hasattr(config, "hidden_sizes"))
self.parent.assertTrue(hasattr(config, "num_attention_heads"))
self.parent.assertTrue(hasattr(config, "num_encoder_blocks"))
class GLPNModelTester:
def __init__(
self,
parent,
batch_size=13,
image_size=64,
num_channels=3,
num_encoder_blocks=4,
depths=[2, 2, 2, 2],
sr_ratios=[8, 4, 2, 1],
hidden_sizes=[16, 32, 64, 128],
downsampling_rates=[1, 4, 8, 16],
num_attention_heads=[1, 2, 4, 8],
is_training=True,
use_labels=True,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
initializer_range=0.02,
decoder_hidden_size=16,
num_labels=3,
scope=None,
):
self.parent = parent
self.batch_size = batch_size
self.image_size = image_size
self.num_channels = num_channels
self.num_encoder_blocks = num_encoder_blocks
self.sr_ratios = sr_ratios
self.depths = depths
self.hidden_sizes = hidden_sizes
self.downsampling_rates = downsampling_rates
self.num_attention_heads = num_attention_heads
self.is_training = is_training
self.use_labels = use_labels
self.hidden_act = hidden_act
self.hidden_dropout_prob = hidden_dropout_prob
self.attention_probs_dropout_prob = attention_probs_dropout_prob
self.initializer_range = initializer_range
self.decoder_hidden_size = decoder_hidden_size
self.num_labels = num_labels
self.scope = scope
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 GLPNConfig(
image_size=self.image_size,
num_channels=self.num_channels,
num_encoder_blocks=self.num_encoder_blocks,
depths=self.depths,
hidden_sizes=self.hidden_sizes,
num_attention_heads=self.num_attention_heads,
hidden_act=self.hidden_act,
hidden_dropout_prob=self.hidden_dropout_prob,
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
initializer_range=self.initializer_range,
decoder_hidden_size=self.decoder_hidden_size,
)
def create_and_check_model(self, config, pixel_values, labels):
model = GLPNModel(config=config)
model.to(torch_device)
model.eval()
result = model(pixel_values)
expected_height = expected_width = self.image_size // (self.downsampling_rates[-1] * 2)
self.parent.assertEqual(
result.last_hidden_state.shape, (self.batch_size, self.hidden_sizes[-1], expected_height, expected_width)
)
def create_and_check_for_depth_estimation(self, config, pixel_values, labels):
config.num_labels = self.num_labels
model = GLPNForDepthEstimation(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))
result = model(pixel_values, labels=labels)
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 GLPNModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
all_model_classes = (GLPNModel, GLPNForDepthEstimation) if is_torch_available() else ()
pipeline_model_mapping = (
{"depth-estimation": GLPNForDepthEstimation, "feature-extraction": GLPNModel} if is_torch_available() else {}
)
test_head_masking = False
test_pruning = False
test_resize_embeddings = False
def setUp(self):
self.model_tester = GLPNModelTester(self)
self.config_tester = GLPNConfigTester(self, config_class=GLPNConfig)
def test_config(self):
self.config_tester.run_common_tests()
def test_model(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*config_and_inputs)
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("GLPN does not use inputs_embeds")
def test_inputs_embeds(self):
pass
@unittest.skip("GLPN does not have get_input_embeddings method and get_output_embeddings methods")
def test_model_common_attributes(self):
pass
def test_forward_signature(self):
config, _ = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
model = model_class(config)
signature = inspect.signature(model.forward)
# signature.parameters is an OrderedDict => so arg_names order is deterministic
arg_names = [*signature.parameters.keys()]
expected_arg_names = ["pixel_values"]
self.assertListEqual(arg_names[:1], expected_arg_names)
def test_attention_outputs(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
config.return_dict = True
for model_class in self.all_model_classes:
inputs_dict["output_attentions"] = True
inputs_dict["output_hidden_states"] = False
config.return_dict = True
model = model_class(config)
model.to(torch_device)
model.eval()
with torch.no_grad():
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
attentions = outputs.attentions
expected_num_attentions = sum(self.model_tester.depths)
self.assertEqual(len(attentions), expected_num_attentions)
# check that output_attentions also work using config
del inputs_dict["output_attentions"]
config.output_attentions = True
model = model_class(config)
model.to(torch_device)
model.eval()
with torch.no_grad():
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
attentions = outputs.attentions
self.assertEqual(len(attentions), expected_num_attentions)
# verify the first attentions (first block, first layer)
expected_seq_len = (self.model_tester.image_size // 4) ** 2
expected_reduced_seq_len = (self.model_tester.image_size // (4 * self.model_tester.sr_ratios[0])) ** 2
self.assertListEqual(
list(attentions[0].shape[-3:]),
[self.model_tester.num_attention_heads[0], expected_seq_len, expected_reduced_seq_len],
)
# verify the last attentions (last block, last layer)
expected_seq_len = (self.model_tester.image_size // 32) ** 2
expected_reduced_seq_len = (self.model_tester.image_size // (32 * self.model_tester.sr_ratios[-1])) ** 2
self.assertListEqual(
list(attentions[-1].shape[-3:]),
[self.model_tester.num_attention_heads[-1], expected_seq_len, expected_reduced_seq_len],
)
out_len = len(outputs)
# Check attention is always last and order is fine
inputs_dict["output_attentions"] = True
inputs_dict["output_hidden_states"] = True
model = model_class(config)
model.to(torch_device)
model.eval()
with torch.no_grad():
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
self.assertEqual(out_len + 1, len(outputs))
self_attentions = outputs.attentions
self.assertEqual(len(self_attentions), expected_num_attentions)
# verify the first attentions (first block, first layer)
expected_seq_len = (self.model_tester.image_size // 4) ** 2
expected_reduced_seq_len = (self.model_tester.image_size // (4 * self.model_tester.sr_ratios[0])) ** 2
self.assertListEqual(
list(self_attentions[0].shape[-3:]),
[self.model_tester.num_attention_heads[0], expected_seq_len, expected_reduced_seq_len],
)
def test_hidden_states_output(self):
def check_hidden_states_output(inputs_dict, config, model_class):
model = model_class(config)
model.to(torch_device)
model.eval()
with torch.no_grad():
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
hidden_states = outputs.hidden_states
expected_num_layers = self.model_tester.num_encoder_blocks
self.assertEqual(len(hidden_states), expected_num_layers)
# verify the first hidden states (first block)
self.assertListEqual(
list(hidden_states[0].shape[-3:]),
[
self.model_tester.hidden_sizes[0],
self.model_tester.image_size // 4,
self.model_tester.image_size // 4,
],
)
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
inputs_dict["output_hidden_states"] = True
check_hidden_states_output(inputs_dict, config, model_class)
# check that output_hidden_states also work using config
del inputs_dict["output_hidden_states"]
config.output_hidden_states = True
check_hidden_states_output(inputs_dict, config, model_class)
def test_training(self):
if not self.model_tester.is_training:
return
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
config.return_dict = True
for model_class in self.all_model_classes:
if model_class in get_values(MODEL_MAPPING):
continue
# TODO: remove the following 3 lines once we have a MODEL_FOR_DEPTH_ESTIMATION_MAPPING
# this can then be incorporated into _prepare_for_class in test_modeling_common.py
if model_class.__name__ == "GLPNForDepthEstimation":
batch_size, num_channels, height, width = inputs_dict["pixel_values"].shape
inputs_dict["labels"] = torch.zeros(
[self.model_tester.batch_size, height, width], device=torch_device
).long()
model = model_class(config)
model.to(torch_device)
model.train()
inputs = self._prepare_for_class(inputs_dict, model_class, return_labels=True)
loss = model(**inputs).loss
loss.backward()
@slow
def test_model_from_pretrained(self):
for model_name in GLPN_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
model = GLPNModel.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 GLPNModelIntegrationTest(unittest.TestCase):
@slow
def test_inference_depth_estimation(self):
image_processor = GLPNImageProcessor.from_pretrained(GLPN_PRETRAINED_MODEL_ARCHIVE_LIST[0])
model = GLPNForDepthEstimation.from_pretrained(GLPN_PRETRAINED_MODEL_ARCHIVE_LIST[0]).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)
# verify the predicted depth
expected_shape = torch.Size([1, 480, 640])
self.assertEqual(outputs.predicted_depth.shape, expected_shape)
expected_slice = torch.tensor(
[[3.4291, 2.7865, 2.5151], [3.2841, 2.7021, 2.3502], [3.1147, 2.4625, 2.2481]]
).to(torch_device)
self.assertTrue(torch.allclose(outputs.predicted_depth[0, :3, :3], expected_slice, atol=1e-4))
|