binary mode consumes a lot of RAM, added non binary mode
Browse files- DUDE_loader.py +38 -33
- test_loader.py +1 -1
DUDE_loader.py
CHANGED
@@ -108,21 +108,38 @@ def open_pdf_binary(pdf_file):
|
|
108 |
return f.read()
|
109 |
|
110 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
111 |
class DUDE(datasets.GeneratorBasedBuilder):
|
112 |
"""DUDE dataset."""
|
113 |
|
|
|
|
|
114 |
BUILDER_CONFIGS = [
|
115 |
-
|
116 |
-
|
117 |
-
version=datasets.Version("0.0.1"),
|
118 |
-
description=_DESCRIPTION,
|
119 |
-
)
|
120 |
]
|
121 |
|
122 |
-
DEFAULT_CONFIG_NAME = "DUDE"
|
123 |
|
124 |
-
def _info(self):
|
125 |
|
|
|
126 |
features = datasets.Features(
|
127 |
{
|
128 |
"docId": datasets.Value("string"),
|
@@ -141,8 +158,8 @@ class DUDE(datasets.GeneratorBasedBuilder):
|
|
141 |
"answers_variants": datasets.Sequence(datasets.Value("string")),
|
142 |
"answer_type": datasets.Value("string"),
|
143 |
"data_split": datasets.Value("string"),
|
144 |
-
"document": datasets.Value("binary"),
|
145 |
-
"OCR": datasets.Value("binary"),
|
146 |
}
|
147 |
)
|
148 |
|
@@ -190,44 +207,32 @@ class DUDE(datasets.GeneratorBasedBuilder):
|
|
190 |
def _generate_examples(self, binary_extraction_path, annotations, split):
|
191 |
def retrieve_doc(docid):
|
192 |
extracted_path = os.path.join(binary_extraction_path, "PDF", split, docid + ".pdf")
|
193 |
-
|
194 |
-
|
195 |
|
196 |
def retrieve_OCR(docid, ocr_engine="Amazon", format="original"):
|
197 |
extracted_path = os.path.join(
|
198 |
binary_extraction_path, "OCR", ocr_engine, docid + f"_{format}.json"
|
199 |
)
|
200 |
-
|
201 |
-
with open(extracted_path, "rb") as f:
|
202 |
-
return f.read()
|
203 |
|
204 |
question = self.info.features["question"]
|
205 |
answers = self.info.features["answers"]
|
206 |
|
207 |
-
extensions = {"pdf", "PDF"}
|
208 |
-
|
209 |
annotations = [x for x in annotations if x["data_split"] == split]
|
210 |
|
211 |
for i, a in enumerate(annotations):
|
212 |
a["data_split"] = split
|
213 |
if a["docId"] in SKIP_DOC_IDS:
|
214 |
continue
|
215 |
-
a["document"] = retrieve_doc(a["docId"])
|
216 |
-
a["OCR"] = retrieve_OCR(a["docId"])
|
217 |
-
|
218 |
a["answers_page_bounding_boxes"] = parse_bbox(a["answers_page_bounding_boxes"])
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
a["answers_page_bounding_boxes"] = None
|
226 |
else:
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
try:
|
231 |
-
except Exception as e:
|
232 |
-
print(f"Something wrong in {split}-{i} {e}")
|
233 |
-
"""
|
|
|
108 |
return f.read()
|
109 |
|
110 |
|
111 |
+
class DUDEConfig(datasets.BuilderConfig):
|
112 |
+
"""BuilderConfig for DUDE."""
|
113 |
+
|
114 |
+
def __init__(
|
115 |
+
self,
|
116 |
+
binary_mode: bool,
|
117 |
+
**kwargs,
|
118 |
+
):
|
119 |
+
"""BuilderConfig for DUDE.
|
120 |
+
Args:
|
121 |
+
binary_mode: `boolean`, load binary PDFs/OCR or pass along paths on local file system
|
122 |
+
**kwargs: keyword arguments forwarded to super.
|
123 |
+
"""
|
124 |
+
BINARY_MODE = False
|
125 |
+
super(DUDEConfig, self).__init__(description=_DESCRIPTION, **kwargs)
|
126 |
+
self.binary_mode = binary_mode or BINARY_MODE
|
127 |
+
|
128 |
+
|
129 |
class DUDE(datasets.GeneratorBasedBuilder):
|
130 |
"""DUDE dataset."""
|
131 |
|
132 |
+
VERSION = datasets.Version("0.0.1")
|
133 |
+
|
134 |
BUILDER_CONFIGS = [
|
135 |
+
DUDEConfig(name='DUDE', version=VERSION, binary_mode=False),
|
136 |
+
DUDEConfig(name='DUDE-binary', version=VERSION, binary_mode=True)
|
|
|
|
|
|
|
137 |
]
|
138 |
|
139 |
+
DEFAULT_CONFIG_NAME = "DUDE" #for some reason not working
|
140 |
|
|
|
141 |
|
142 |
+
def _info(self):
|
143 |
features = datasets.Features(
|
144 |
{
|
145 |
"docId": datasets.Value("string"),
|
|
|
158 |
"answers_variants": datasets.Sequence(datasets.Value("string")),
|
159 |
"answer_type": datasets.Value("string"),
|
160 |
"data_split": datasets.Value("string"),
|
161 |
+
"document": datasets.Value("binary") if self.config.binary_mode else datasets.Value("string"),
|
162 |
+
"OCR": datasets.Value("binary") if self.config.binary_mode else datasets.Value("string"),
|
163 |
}
|
164 |
)
|
165 |
|
|
|
207 |
def _generate_examples(self, binary_extraction_path, annotations, split):
|
208 |
def retrieve_doc(docid):
|
209 |
extracted_path = os.path.join(binary_extraction_path, "PDF", split, docid + ".pdf")
|
210 |
+
return extracted_path
|
211 |
+
|
212 |
|
213 |
def retrieve_OCR(docid, ocr_engine="Amazon", format="original"):
|
214 |
extracted_path = os.path.join(
|
215 |
binary_extraction_path, "OCR", ocr_engine, docid + f"_{format}.json"
|
216 |
)
|
217 |
+
return extracted_path
|
|
|
|
|
218 |
|
219 |
question = self.info.features["question"]
|
220 |
answers = self.info.features["answers"]
|
221 |
|
|
|
|
|
222 |
annotations = [x for x in annotations if x["data_split"] == split]
|
223 |
|
224 |
for i, a in enumerate(annotations):
|
225 |
a["data_split"] = split
|
226 |
if a["docId"] in SKIP_DOC_IDS:
|
227 |
continue
|
|
|
|
|
|
|
228 |
a["answers_page_bounding_boxes"] = parse_bbox(a["answers_page_bounding_boxes"])
|
229 |
+
docpath = retrieve_doc(a["docId"])
|
230 |
+
ocrpath = retrieve_OCR(a["docId"])
|
231 |
+
if self.config.binary_mode:
|
232 |
+
with open(docpath, "rb") as f, open(ocrpath, "rb") as g:
|
233 |
+
a["document"] = f.read()
|
234 |
+
a["OCR"] = g.read()
|
|
|
235 |
else:
|
236 |
+
a["document"] = docpath
|
237 |
+
a["OCR"] = ocrpath
|
238 |
+
yield i, a
|
|
|
|
|
|
|
|
test_loader.py
CHANGED
@@ -22,7 +22,7 @@ from codetiming import Timer
|
|
22 |
for binding in ["dict_annotations (new)"]: #"dict_PDF",
|
23 |
with Timer(name=f"{binding}", text=binding + " Elapsed time: {:.4f} seconds"):
|
24 |
if binding == "dict_annotations (new)":
|
25 |
-
ds = load_dataset("../DUDE_loader/DUDE_loader.py", data_dir="/home/jordy/Downloads/DUDE_train-val-test_binaries"
|
26 |
else:
|
27 |
ds = load_dataset("jordyvl/DUDE_loader", revision='db20bbf751b14e14e8143170bc201948ef5ac83c')
|
28 |
|
|
|
22 |
for binding in ["dict_annotations (new)"]: #"dict_PDF",
|
23 |
with Timer(name=f"{binding}", text=binding + " Elapsed time: {:.4f} seconds"):
|
24 |
if binding == "dict_annotations (new)":
|
25 |
+
ds = load_dataset("../DUDE_loader/DUDE_loader.py", 'DUDE', data_dir="/home/jordy/Downloads/DUDE_train-val-test_binaries") #ignore_verifications=True, , writer_batch_size=10
|
26 |
else:
|
27 |
ds = load_dataset("jordyvl/DUDE_loader", revision='db20bbf751b14e14e8143170bc201948ef5ac83c')
|
28 |
|