changed to input ids
Browse files- ref_seg_ger.py +29 -24
ref_seg_ger.py
CHANGED
@@ -18,7 +18,7 @@ from glob import glob
|
|
18 |
import os
|
19 |
import numpy as np
|
20 |
from PIL import Image
|
21 |
-
from
|
22 |
import datasets
|
23 |
from itertools import chain
|
24 |
import pandas as pd
|
@@ -62,7 +62,7 @@ _LABELS = [
|
|
62 |
_FEATURES = datasets.Features(
|
63 |
{
|
64 |
#"id": datasets.Value("string"),
|
65 |
-
"input_ids": datasets.Sequence(datasets.Value("
|
66 |
#"bbox": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
|
67 |
# "RGBs": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
|
68 |
# "fonts": datasets.Sequence(datasets.Value("string")),
|
@@ -136,7 +136,7 @@ class RefSeg(datasets.GeneratorBasedBuilder):
|
|
136 |
# ]
|
137 |
|
138 |
# DEFAULT_CONFIG_NAME = "small" # It's not mandatory to have a default configuration. Just use one if it make sense.
|
139 |
-
TOKENIZER =
|
140 |
|
141 |
def _info(self):
|
142 |
# TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
|
@@ -214,25 +214,30 @@ class RefSeg(datasets.GeneratorBasedBuilder):
|
|
214 |
labels = []
|
215 |
for i, row in df.iterrows():
|
216 |
|
217 |
-
|
|
|
|
|
|
|
|
|
|
|
218 |
#print(tokenized_input)
|
219 |
-
if len(
|
220 |
if row['tag'] == 'B':
|
221 |
-
input_ids.append(
|
222 |
labels.append(row['tag'] + '-' + row['label'])
|
223 |
-
for input_id in
|
224 |
-
input_ids.append(input_id
|
225 |
labels.append('I-' + row['label'])
|
226 |
elif row['tag'] == 'I':
|
227 |
-
for input_id in
|
228 |
-
input_ids.append(input_id
|
229 |
labels.append('I-' + row['label'])
|
230 |
else:
|
231 |
-
for input_id in
|
232 |
-
input_ids.append(input_id
|
233 |
labels.append('O')
|
234 |
-
elif len(
|
235 |
-
input_ids.append(
|
236 |
if row['tag'] == 'O':
|
237 |
labels.append(row['tag'])
|
238 |
else:
|
@@ -247,16 +252,16 @@ class RefSeg(datasets.GeneratorBasedBuilder):
|
|
247 |
# split_rgbs = rgbs[index:index + self.CHUNK_SIZE]
|
248 |
# split_fonts = fonts[index:index + self.CHUNK_SIZE]
|
249 |
split_labels = labels[index:max(len(input_ids), index + self.CHUNK_SIZE)]
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
#print(len(split_labels_post), split_labels_post)
|
261 |
#print(split_labels, len(split_labels))
|
262 |
#print(split_ids, len(split_ids))
|
|
|
18 |
import os
|
19 |
import numpy as np
|
20 |
from PIL import Image
|
21 |
+
from transformers import AutoTokenizer
|
22 |
import datasets
|
23 |
from itertools import chain
|
24 |
import pandas as pd
|
|
|
62 |
_FEATURES = datasets.Features(
|
63 |
{
|
64 |
#"id": datasets.Value("string"),
|
65 |
+
"input_ids": datasets.Sequence(datasets.Value("int64")),
|
66 |
#"bbox": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
|
67 |
# "RGBs": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
|
68 |
# "fonts": datasets.Sequence(datasets.Value("string")),
|
|
|
136 |
# ]
|
137 |
|
138 |
# DEFAULT_CONFIG_NAME = "small" # It's not mandatory to have a default configuration. Just use one if it make sense.
|
139 |
+
TOKENIZER = AutoTokenizer.from_pretrained("xlm-roberta-base")
|
140 |
|
141 |
def _info(self):
|
142 |
# TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
|
|
|
214 |
labels = []
|
215 |
for i, row in df.iterrows():
|
216 |
|
217 |
+
tokenized_input = self.TOKENIZER(
|
218 |
+
row['token'],
|
219 |
+
add_special_tokens=False,
|
220 |
+
return_offsets_mapping=False,
|
221 |
+
return_attention_mask=False,
|
222 |
+
)
|
223 |
#print(tokenized_input)
|
224 |
+
if len(tokenized_input['input_ids']) > 1:
|
225 |
if row['tag'] == 'B':
|
226 |
+
input_ids.append(tokenized_input['input_ids'][0])
|
227 |
labels.append(row['tag'] + '-' + row['label'])
|
228 |
+
for input_id in tokenized_input['input_ids'][1:]:
|
229 |
+
input_ids.append(input_id)
|
230 |
labels.append('I-' + row['label'])
|
231 |
elif row['tag'] == 'I':
|
232 |
+
for input_id in tokenized_input['input_ids']:
|
233 |
+
input_ids.append(input_id)
|
234 |
labels.append('I-' + row['label'])
|
235 |
else:
|
236 |
+
for input_id in tokenized_input['input_ids']:
|
237 |
+
input_ids.append(input_id)
|
238 |
labels.append('O')
|
239 |
+
elif len(tokenized_input['input_ids']) == 1:
|
240 |
+
input_ids.append(tokenized_input['input_ids'][0])
|
241 |
if row['tag'] == 'O':
|
242 |
labels.append(row['tag'])
|
243 |
else:
|
|
|
252 |
# split_rgbs = rgbs[index:index + self.CHUNK_SIZE]
|
253 |
# split_fonts = fonts[index:index + self.CHUNK_SIZE]
|
254 |
split_labels = labels[index:max(len(input_ids), index + self.CHUNK_SIZE)]
|
255 |
+
split_labels_post = [item for sublist in split_labels for item in sublist]
|
256 |
+
if(len(split_ids) != len(split_labels)):
|
257 |
+
print(f)
|
258 |
+
print(len(input_ids), input_ids)
|
259 |
+
print(len(split_labels), split_labels)
|
260 |
+
for s in split_labels:
|
261 |
+
if type(s) is not str:
|
262 |
+
print(f)
|
263 |
+
print(len(input_ids), input_ids)
|
264 |
+
print(len(split_labels), split_labels)
|
265 |
#print(len(split_labels_post), split_labels_post)
|
266 |
#print(split_labels, len(split_labels))
|
267 |
#print(split_ids, len(split_ids))
|