File size: 11,119 Bytes
e857f97 | 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 | import argparse
import torch
from torch.utils.data import DataLoader
from data.build_datasets import get_data
from .dataloader_msrvtt_retrieval import MSRVTT_DataLoader
from .dataloader_msrvtt_retrieval import MSRVTT_TrainDataLoader
from .dataloader_msvd_retrieval import MSVD_DataLoader
from .dataloader_lsmdc_retrieval import LSMDC_DataLoader
from .dataloader_activitynet_retrieval import ActivityNet_DataLoader
from .dataloader_didemo_retrieval import DiDeMo_DataLoader
def dataloader_msrvtt_train(args, tokenizer):
msrvtt_dataset = MSRVTT_TrainDataLoader(
csv_path=args.train_csv,
json_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
unfold_sentences=args.expand_msrvtt_sentences,
frame_order=args.train_frame_order,
slice_framepos=args.slice_framepos,
)
train_sampler = torch.utils.data.distributed.DistributedSampler(msrvtt_dataset)
dataloader = DataLoader(
msrvtt_dataset,
batch_size=args.batch_size // args.n_gpu,
num_workers=args.num_thread_reader,
pin_memory=False,
shuffle=(train_sampler is None),
sampler=train_sampler,
drop_last=True,
)
return dataloader, len(msrvtt_dataset), train_sampler
def dataloader_msrvtt_test(args, tokenizer, subset="test"):
msrvtt_testset = MSRVTT_DataLoader(
csv_path=args.val_csv,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.eval_frame_order,
slice_framepos=args.slice_framepos,
)
dataloader_msrvtt = DataLoader(
msrvtt_testset,
batch_size=args.batch_size_val,
num_workers=args.num_thread_reader,
shuffle=False,
drop_last=False,
)
return dataloader_msrvtt, len(msrvtt_testset)
def dataloader_msvd_train(args, tokenizer):
msvd_dataset = MSVD_DataLoader(
subset="train",
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.train_frame_order,
slice_framepos=args.slice_framepos,
)
train_sampler = torch.utils.data.distributed.DistributedSampler(msvd_dataset)
dataloader = DataLoader(
msvd_dataset,
batch_size=args.batch_size // args.n_gpu,
num_workers=args.num_thread_reader,
pin_memory=False,
shuffle=(train_sampler is None),
sampler=train_sampler,
drop_last=True,
)
return dataloader, len(msvd_dataset), train_sampler
def dataloader_msvd_test(args, tokenizer, subset="test"):
msvd_testset = MSVD_DataLoader(
subset=subset,
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.eval_frame_order,
slice_framepos=args.slice_framepos,
)
dataloader_msrvtt = DataLoader(
msvd_testset,
batch_size=args.batch_size_val,
num_workers=args.num_thread_reader,
shuffle=False,
drop_last=False,
)
return dataloader_msrvtt, len(msvd_testset)
def dataloader_lsmdc_train(args, tokenizer):
lsmdc_dataset = LSMDC_DataLoader(
subset="train",
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.train_frame_order,
slice_framepos=args.slice_framepos,
)
train_sampler = torch.utils.data.distributed.DistributedSampler(lsmdc_dataset)
dataloader = DataLoader(
lsmdc_dataset,
batch_size=args.batch_size // args.n_gpu,
num_workers=args.num_thread_reader,
pin_memory=False,
shuffle=(train_sampler is None),
sampler=train_sampler,
drop_last=True,
)
return dataloader, len(lsmdc_dataset), train_sampler
def dataloader_lsmdc_test(args, tokenizer, subset="test"):
lsmdc_testset = LSMDC_DataLoader(
subset=subset,
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.eval_frame_order,
slice_framepos=args.slice_framepos,
)
dataloader_msrvtt = DataLoader(
lsmdc_testset,
batch_size=args.batch_size_val,
num_workers=args.num_thread_reader,
shuffle=False,
drop_last=False,
)
return dataloader_msrvtt, len(lsmdc_testset)
def dataloader_activity_train(args, tokenizer):
activity_dataset = ActivityNet_DataLoader(
subset="train",
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.train_frame_order,
slice_framepos=args.slice_framepos,
)
train_sampler = torch.utils.data.distributed.DistributedSampler(activity_dataset)
dataloader = DataLoader(
activity_dataset,
batch_size=args.batch_size // args.n_gpu,
num_workers=args.num_thread_reader,
pin_memory=False,
shuffle=(train_sampler is None),
sampler=train_sampler,
drop_last=True,
)
return dataloader, len(activity_dataset), train_sampler
def dataloader_activity_test(args, tokenizer, subset="test"):
activity_testset = ActivityNet_DataLoader(
subset=subset,
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.eval_frame_order,
slice_framepos=args.slice_framepos,
)
dataloader_msrvtt = DataLoader(
activity_testset,
batch_size=args.batch_size_val,
num_workers=args.num_thread_reader,
shuffle=False,
drop_last=False,
)
return dataloader_msrvtt, len(activity_testset)
def dataloader_didemo_train(args, tokenizer):
didemo_dataset = DiDeMo_DataLoader(
subset="train",
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.train_frame_order,
slice_framepos=args.slice_framepos,
)
train_sampler = torch.utils.data.distributed.DistributedSampler(didemo_dataset)
dataloader = DataLoader(
didemo_dataset,
batch_size=args.batch_size // args.n_gpu,
num_workers=args.num_thread_reader,
pin_memory=False,
shuffle=(train_sampler is None),
sampler=train_sampler,
drop_last=True,
)
return dataloader, len(didemo_dataset), train_sampler
def dataloader_didemo_test(args, tokenizer, subset="test"):
didemo_testset = DiDeMo_DataLoader(
subset=subset,
data_path=args.data_path,
features_path=args.features_path,
max_words=args.max_words,
feature_framerate=args.feature_framerate,
tokenizer=tokenizer,
max_frames=args.max_frames,
frame_order=args.eval_frame_order,
slice_framepos=args.slice_framepos,
)
dataloader_didemo = DataLoader(
didemo_testset,
batch_size=args.batch_size_val,
num_workers=args.num_thread_reader,
shuffle=False,
drop_last=False,
)
return dataloader_didemo, len(didemo_testset)
DATALOADER_DICT = {}
DATALOADER_DICT["msrvtt"] = {"train":dataloader_msrvtt_train, "val":dataloader_msrvtt_test, "test":None}
DATALOADER_DICT["msvd"] = {"train":dataloader_msvd_train, "val":dataloader_msvd_test, "test":dataloader_msvd_test}
DATALOADER_DICT["lsmdc"] = {"train":dataloader_lsmdc_train, "val":dataloader_lsmdc_test, "test":dataloader_lsmdc_test}
DATALOADER_DICT["activity"] = {"train":dataloader_activity_train, "val":dataloader_activity_test, "test":None}
DATALOADER_DICT["didemo"] = {"train":dataloader_didemo_train, "val":dataloader_didemo_test, "test":dataloader_didemo_test}
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("--val_vl_ret_data", default="", type=str, help="Point the dataset to finetune.")
parser.add_argument("--val_v_cls_data", default="", type=str, help="Point the dataset to finetune.")
parser.add_argument("--do_train", action='store_true', help="Whether to run training.")
parser.add_argument("--do_eval", action='store_true', help="Whether to run eval on the dev set.")
parser.add_argument('--train_csv', type=str, default='data/.train.csv', help='')
parser.add_argument('--val_csv', type=str, default='data/.val.csv', help='')
parser.add_argument('--data_path', type=str, default='data/caption.pickle', help='data pickle file path')
parser.add_argument('--features_path', type=str, default='data/videos_feature.pickle', help='feature path')
parser.add_argument('--eval_frame_order', type=int, default=0, choices=[0, 1, 2],
help="Frame order, 0: ordinary order; 1: reverse order; 2: random order.")
parser.add_argument('--feature_framerate', type=int, default=1, help='')
parser.add_argument('--slice_framepos', type=int, default=2, choices=[0, 1, 2],
help="0: cut from head frames; 1: cut from tail frames; 2: extract frames uniformly.")
parser.add_argument('--max_frames', type=int, default=100, help='')
parser.add_argument('--max_words', type=int, default=20, help='')
parser.add_argument('--batch_size', type=int, default=77, help='')
parser.add_argument('--workers', type=int, default=0, help='')
parser.add_argument('--batch_size_val', type=int, default=0, help='batch size eval')
parser.add_argument('--num_thread_reader', type=int, default=0, help='')
parser.add_argument('--num_frames', type=int, default=8, help='')
args = parser.parse_args()
args.val_vl_ret_data = 'msrvtt'
args.do_train = False
args.do_eval = True
args.slice_framepos = 2
args.max_words = 77
args.train_csv = 'D:/MSRVTT/MSRVTT_train.9k.csv'
args.val_csv = 'D:/MSRVTT/MSRVTT_JSFUSION_test.csv'
args.data_path = 'D:/MSRVTT/MSRVTT_data.json'
args.features_path = 'D:/MSRVTT/videos/all'
dataloader_msrvtt = get_data(args)["vl_ret"]
for batch in dataloader_msrvtt:
print() |