try using manual dir
Browse files- SNLI-VE.py +7 -6
SNLI-VE.py
CHANGED
@@ -4,6 +4,7 @@
|
|
4 |
import json
|
5 |
import os
|
6 |
import datasets
|
|
|
7 |
|
8 |
_CITATION = """\
|
9 |
@article{xie2019visual,
|
@@ -59,7 +60,6 @@ _SNLI_VE_SPLITS = {
|
|
59 |
"validation": "snli_ve_dev.jsonl",
|
60 |
"test": "snli_ve_test.jsonl",
|
61 |
}
|
62 |
-
JZ_FOLDER_PATH = f"{os.environ['cnw_ALL_CCFRSCRATCH']}/local_datasets/flickr30k-images.tar.gz"
|
63 |
|
64 |
_FEATURES = datasets.Features(
|
65 |
{
|
@@ -75,7 +75,7 @@ _FEATURES = datasets.Features(
|
|
75 |
class SNLIVE(datasets.GeneratorBasedBuilder):
|
76 |
"""SNLIVE."""
|
77 |
|
78 |
-
|
79 |
DEFAULT_CONFIG_NAME = "Default"
|
80 |
def _info(self):
|
81 |
return datasets.DatasetInfo(
|
@@ -87,6 +87,11 @@ class SNLIVE(datasets.GeneratorBasedBuilder):
|
|
87 |
)
|
88 |
|
89 |
def _split_generators(self, dl_manager):
|
|
|
|
|
|
|
|
|
|
|
90 |
urls = {
|
91 |
"Default": {
|
92 |
"train": os.path.join(_SNLI_VE_URL_BASE, _SNLI_VE_SPLITS["train"]),
|
@@ -95,10 +100,6 @@ class SNLIVE(datasets.GeneratorBasedBuilder):
|
|
95 |
},
|
96 |
}
|
97 |
snli_ve_annotation_path = dl_manager.download_and_extract(urls)
|
98 |
-
images_path = os.path.join(
|
99 |
-
dl_manager.extract(JZ_FOLDER_PATH),
|
100 |
-
"flickr30k-images"
|
101 |
-
)
|
102 |
|
103 |
return [
|
104 |
datasets.SplitGenerator(
|
|
|
4 |
import json
|
5 |
import os
|
6 |
import datasets
|
7 |
+
from pathlib import Path
|
8 |
|
9 |
_CITATION = """\
|
10 |
@article{xie2019visual,
|
|
|
60 |
"validation": "snli_ve_dev.jsonl",
|
61 |
"test": "snli_ve_test.jsonl",
|
62 |
}
|
|
|
63 |
|
64 |
_FEATURES = datasets.Features(
|
65 |
{
|
|
|
75 |
class SNLIVE(datasets.GeneratorBasedBuilder):
|
76 |
"""SNLIVE."""
|
77 |
|
78 |
+
_LOCAL_IMAGE_FOLDER_NAME = "flickr30k-images"
|
79 |
DEFAULT_CONFIG_NAME = "Default"
|
80 |
def _info(self):
|
81 |
return datasets.DatasetInfo(
|
|
|
87 |
)
|
88 |
|
89 |
def _split_generators(self, dl_manager):
|
90 |
+
images_path = os.path.join(
|
91 |
+
dl_manager.extract(os.path.join(dl_manager.manual_dir, "flickr30k-images.tar.gz")),
|
92 |
+
self._LOCAL_IMAGE_FOLDER_NAME
|
93 |
+
)
|
94 |
+
|
95 |
urls = {
|
96 |
"Default": {
|
97 |
"train": os.path.join(_SNLI_VE_URL_BASE, _SNLI_VE_SPLITS["train"]),
|
|
|
100 |
},
|
101 |
}
|
102 |
snli_ve_annotation_path = dl_manager.download_and_extract(urls)
|
|
|
|
|
|
|
|
|
103 |
|
104 |
return [
|
105 |
datasets.SplitGenerator(
|