def _download(self, annFile):
"""
Utility function for downloading the COCO annotation file
:param annFile: path of the annotations file
:return: void - extracts the archive
"""
if not os.path.exists(annFile):
if not os.path.isdir(annFile):
if "2017" in annFile:
annotations_dir_zip = os.path.join(
self.root, "annotations_train%s2017.zip" % self.split
)
elif "2014" in annFile:
annotations_dir_zip = os.path.join(
self.root, "annotations_train%s2014.zip" % self.split
)
else:
annotations_dir_zip = None
if annotations_dir_zip is not None:
print(
"Attempt to extract annotations file at {zip_loc}".format(
zip_loc=annotations_dir_zip
)
)
extract_archive(from_path=annotations_dir_zip, to_path=self.root)
def _get_path(self, local_root, local_unzip=False):
root = Path(change_root_if_server(root=local_root,
server_root=".data/nlp/" + self.pwc_name.lower()))
zip_name = self.pwc_name.lower() + "-v1.zip"
dataset_path = root / "wiki.test.tokens"
if not dataset_path.exists(): # unzip
extract_archive(str(root / zip_name), to_path=root.parent)
return dataset_path
def get_path(local_root, local_unzip=False):
root = Path(
change_root_if_server(root=local_root,
server_root=".data/nlp/multinli"))
zip_name = "MNLI.zip"
dataset_path = root / "MNLI" / "dev_matched.tsv"
if not dataset_path.exists(): # unzip
extract_archive(str(root / zip_name), to_path=root)
return (dataset_path, dataset_path.parent / "dev_mismatched.tsv")
def _download(self, annFile):
if not os.path.isdir(annFile):
if "2017" in annFile:
annotations_dir_zip = os.path.join(
self.root, "annotations_train%s2017.zip" % self.split)
elif "2014" in annFile:
annotations_dir_zip = os.path.join(
self.root, "annotations_train%s2014.zip" % self.split)
else:
annotations_dir_zip = None
if annotations_dir_zip is not None:
print(
'Attempt to extract annotations file at {zip_loc}'.format(
zip_loc=annotations_dir_zip))
extract_archive(from_path=annotations_dir_zip,
to_path=self.root)
has_amp = True
except ImportError:
has_amp = False
from sotabencheval.object_detection import COCOEvaluator
from sotabencheval.utils import is_server, extract_archive
from effdet import create_model
from data import CocoDetection, create_loader
NUM_GPU = 1
BATCH_SIZE = (128 if has_amp else 64) * NUM_GPU
ANNO_SET = 'val2017'
if is_server():
DATA_ROOT = './.data/vision/coco'
image_dir_zip = os.path.join('./.data/vision/coco', f'{ANNO_SET}.zip')
extract_archive(from_path=image_dir_zip, to_path='./.data/vision/coco')
else:
# local settings
DATA_ROOT = ''
def _bs(b=64):
b *= NUM_GPU
if has_amp:
b *= 2
return b
def _entry(model_name, paper_model_name, paper_arxiv_id, batch_size=BATCH_SIZE, model_desc=None):
return dict(
model_name=model_name,