Spaces:
Paused
Paused
File size: 13,383 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 |
# Copyright 2022 The HuggingFace 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.
import unittest
from transformers import MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING, AutoTokenizer, is_vision_available
from transformers.pipelines import pipeline
from transformers.pipelines.document_question_answering import apply_tesseract
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_detectron2,
require_pytesseract,
require_tf,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_vision_available():
from PIL import Image
from transformers.image_utils import load_image
else:
class Image:
@staticmethod
def open(*args, **kwargs):
pass
def load_image(_):
return None
# This is a pinned image from a specific revision of a document question answering space, hosted by HuggingFace,
# so we can expect it to be available.
INVOICE_URL = (
"https://huggingface.co/spaces/impira/docquery/resolve/2f6c96314dc84dfda62d40de9da55f2f5165d403/invoice.png"
)
@is_pipeline_test
@require_torch
@require_vision
class DocumentQuestionAnsweringPipelineTests(unittest.TestCase):
model_mapping = MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING
@require_pytesseract
@require_vision
def get_test_pipeline(self, model, tokenizer, processor):
dqa_pipeline = pipeline(
"document-question-answering", model=model, tokenizer=tokenizer, image_processor=processor
)
image = INVOICE_URL
word_boxes = list(zip(*apply_tesseract(load_image(image), None, "")))
question = "What is the placebo?"
examples = [
{
"image": load_image(image),
"question": question,
},
{
"image": image,
"question": question,
},
{
"image": image,
"question": question,
"word_boxes": word_boxes,
},
]
return dqa_pipeline, examples
def run_pipeline_test(self, dqa_pipeline, examples):
outputs = dqa_pipeline(examples, top_k=2)
self.assertEqual(
outputs,
[
[
{"score": ANY(float), "answer": ANY(str), "start": ANY(int), "end": ANY(int)},
{"score": ANY(float), "answer": ANY(str), "start": ANY(int), "end": ANY(int)},
]
]
* 3,
)
@require_torch
@require_detectron2
@require_pytesseract
def test_small_model_pt(self):
dqa_pipeline = pipeline("document-question-answering", model="hf-internal-testing/tiny-random-layoutlmv2")
image = INVOICE_URL
question = "How many cats are there?"
expected_output = [
{"score": 0.0001, "answer": "oy 2312/2019", "start": 38, "end": 39},
{"score": 0.0001, "answer": "oy 2312/2019 DUE", "start": 38, "end": 40},
]
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(nested_simplify(outputs, decimals=4), expected_output)
outputs = dqa_pipeline({"image": image, "question": question}, top_k=2)
self.assertEqual(nested_simplify(outputs, decimals=4), expected_output)
# This image does not detect ANY text in it, meaning layoutlmv2 should fail.
# Empty answer probably
image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(outputs, [])
# We can optionnally pass directly the words and bounding boxes
image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
words = []
boxes = []
outputs = dqa_pipeline(image=image, question=question, words=words, boxes=boxes, top_k=2)
self.assertEqual(outputs, [])
# TODO: Enable this once hf-internal-testing/tiny-random-donut is implemented
# @require_torch
# def test_small_model_pt_donut(self):
# dqa_pipeline = pipeline("document-question-answering", model="hf-internal-testing/tiny-random-donut")
# # dqa_pipeline = pipeline("document-question-answering", model="../tiny-random-donut")
# image = "https://templates.invoicehome.com/invoice-template-us-neat-750px.png"
# question = "How many cats are there?"
#
# outputs = dqa_pipeline(image=image, question=question, top_k=2)
# self.assertEqual(
# nested_simplify(outputs, decimals=4), [{"score": 0.8799, "answer": "2"}, {"score": 0.296, "answer": "1"}]
# )
@slow
@require_torch
@require_detectron2
@require_pytesseract
def test_large_model_pt(self):
dqa_pipeline = pipeline(
"document-question-answering",
model="tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa",
revision="9977165",
)
image = INVOICE_URL
question = "What is the invoice number?"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9944, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0009, "answer": "us-001", "start": 16, "end": 16},
],
)
outputs = dqa_pipeline({"image": image, "question": question}, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9944, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0009, "answer": "us-001", "start": 16, "end": 16},
],
)
outputs = dqa_pipeline(
[{"image": image, "question": question}, {"image": image, "question": question}], top_k=2
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.9944, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0009, "answer": "us-001", "start": 16, "end": 16},
],
]
* 2,
)
@slow
@require_torch
@require_detectron2
@require_pytesseract
def test_large_model_pt_chunk(self):
dqa_pipeline = pipeline(
"document-question-answering",
model="tiennvcs/layoutlmv2-base-uncased-finetuned-docvqa",
revision="9977165",
max_seq_len=50,
)
image = INVOICE_URL
question = "What is the invoice number?"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9974, "answer": "1110212019", "start": 23, "end": 23},
{"score": 0.9948, "answer": "us-001", "start": 16, "end": 16},
],
)
outputs = dqa_pipeline({"image": image, "question": question}, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9974, "answer": "1110212019", "start": 23, "end": 23},
{"score": 0.9948, "answer": "us-001", "start": 16, "end": 16},
],
)
outputs = dqa_pipeline(
[{"image": image, "question": question}, {"image": image, "question": question}], top_k=2
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.9974, "answer": "1110212019", "start": 23, "end": 23},
{"score": 0.9948, "answer": "us-001", "start": 16, "end": 16},
]
]
* 2,
)
@slow
@require_torch
@require_pytesseract
@require_vision
def test_large_model_pt_layoutlm(self):
tokenizer = AutoTokenizer.from_pretrained(
"impira/layoutlm-document-qa", revision="3dc6de3", add_prefix_space=True
)
dqa_pipeline = pipeline(
"document-question-answering",
model="impira/layoutlm-document-qa",
tokenizer=tokenizer,
revision="3dc6de3",
)
image = INVOICE_URL
question = "What is the invoice number?"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.4251, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0819, "answer": "1110212019", "start": 23, "end": 23},
],
)
outputs = dqa_pipeline({"image": image, "question": question}, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.4251, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0819, "answer": "1110212019", "start": 23, "end": 23},
],
)
outputs = dqa_pipeline(
[{"image": image, "question": question}, {"image": image, "question": question}], top_k=2
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.4251, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0819, "answer": "1110212019", "start": 23, "end": 23},
]
]
* 2,
)
word_boxes = list(zip(*apply_tesseract(load_image(image), None, "")))
# This model should also work if `image` is set to None
outputs = dqa_pipeline({"image": None, "word_boxes": word_boxes, "question": question}, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.4251, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.0819, "answer": "1110212019", "start": 23, "end": 23},
],
)
@slow
@require_torch
@require_pytesseract
@require_vision
def test_large_model_pt_layoutlm_chunk(self):
tokenizer = AutoTokenizer.from_pretrained(
"impira/layoutlm-document-qa", revision="3dc6de3", add_prefix_space=True
)
dqa_pipeline = pipeline(
"document-question-answering",
model="impira/layoutlm-document-qa",
tokenizer=tokenizer,
revision="3dc6de3",
max_seq_len=50,
)
image = INVOICE_URL
question = "What is the invoice number?"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9999, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.9998, "answer": "us-001", "start": 16, "end": 16},
],
)
outputs = dqa_pipeline(
[{"image": image, "question": question}, {"image": image, "question": question}], top_k=2
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.9999, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.9998, "answer": "us-001", "start": 16, "end": 16},
]
]
* 2,
)
word_boxes = list(zip(*apply_tesseract(load_image(image), None, "")))
# This model should also work if `image` is set to None
outputs = dqa_pipeline({"image": None, "word_boxes": word_boxes, "question": question}, top_k=2)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
{"score": 0.9999, "answer": "us-001", "start": 16, "end": 16},
{"score": 0.9998, "answer": "us-001", "start": 16, "end": 16},
],
)
@slow
@require_torch
def test_large_model_pt_donut(self):
dqa_pipeline = pipeline(
"document-question-answering",
model="naver-clova-ix/donut-base-finetuned-docvqa",
tokenizer=AutoTokenizer.from_pretrained("naver-clova-ix/donut-base-finetuned-docvqa"),
feature_extractor="naver-clova-ix/donut-base-finetuned-docvqa",
)
image = INVOICE_URL
question = "What is the invoice number?"
outputs = dqa_pipeline(image=image, question=question, top_k=2)
self.assertEqual(nested_simplify(outputs, decimals=4), [{"answer": "us-001"}])
@require_tf
@unittest.skip("Document question answering not implemented in TF")
def test_small_model_tf(self):
pass
|