def test_train_iter2():
cfg = get_cfg()
cfg.LOG_PERIOD = 1
train_meter = ClevrerTrainMeter(5, cfg)
train_meter.update_stats(top1_err=40.0,
top5_err=20.0,
mc_opt_err=55.0,
mc_q_err=78.0,
loss_des=6.3,
loss_mc=0.2,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
train_meter.update_stats(top1_err=41.0,
top5_err=21.0,
mc_opt_err=56.0,
mc_q_err=86.0,
loss_des=6.7,
loss_mc=1.2,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
stats = train_meter.log_iter_stats(cur_epoch=1, cur_iter=1)
assert stats['top1_err'] == 41.0
assert stats['top5_err'] == 21.0
assert stats['mc_opt_err'] == 56.0
assert stats['mc_q_err'] == 86.0
assert stats['loss_des'] == 6.7
assert stats['loss_mc'] == 1.2
def load_config(args):
"""
Given the arguemnts, load and initialize the configs.
Args:
args (argument): arguments includes `shard_id`, `num_shards`,
`init_method`, `cfg_file`, and `opts`.
"""
# Setup cfg.
cfg = get_cfg()
# Load config from cfg.
if args.cfg_file is not None:
cfg.merge_from_file(args.cfg_file)
# Load config from command line, overwrite config from opts.
if args.opts is not None:
cfg.merge_from_list(args.opts)
# Inherit parameters from args.
if hasattr(args, "num_shards") and hasattr(args, "shard_id"):
cfg.NUM_SHARDS = args.num_shards
cfg.SHARD_ID = args.shard_id
if hasattr(args, "rng_seed"):
cfg.RNG_SEED = args.rng_seed
if hasattr(args, "output_dir"):
cfg.OUTPUT_DIR = args.output_dir
cu.make_checkpoint_dir(cfg.OUTPUT_DIR)
return cfg
def build_teacher_model(cfg, gpu_id=None):
"""
Builds the video model.
Args:
cfg (configs): configs that contains the hyper-parameters to build the
backbone. Details can be seen in slowfast/config/defaults.py.
"""
# Load teacher cfg
teacher_cfg = get_cfg()
teacher_cfg.merge_from_file(cfg.KD.CONFIG)
teacher_cfg.KD.ENABLE = True
# Construct the model
model = MODEL_REGISTRY.get(teacher_cfg.MODEL.MODEL_NAME)(teacher_cfg)
if cfg.NUM_GPUS:
if gpu_id is None:
# Determine the GPU used by the current process
cur_device = torch.cuda.current_device()
else:
cur_device = gpu_id
# Transfer the model to the current GPU device
model = model.cuda(device=cur_device)
# Use multi-process data parallel model in the multi-gpu setting
if cfg.NUM_GPUS > 1:
# Make model replica operate on the current device
model = torch.nn.parallel.DistributedDataParallel(
module=model, device_ids=[cur_device], output_device=cur_device
)
return model
def get_init_params_cfg():
#Init parameters
cfg = get_cfg()
cfg.TRAIN.ENABLE = True
cfg.TRAIN.ONLY_DES = True
cfg.TRAIN.DATASET = "Clevrer_des"
cfg.TRAIN.BATCH_SIZE = 8
cfg.TRAIN.EVAL_PERIOD = 1
cfg.TRAIN.CHECKPOINT_PERIOD = 1
cfg.TRAIN.AUTO_RESUME = True
cfg.TRAIN.TRAIN_STATS_FILE = "./train_stats_hyper.txt"
cfg.DATA.RESIZE_H = 224
cfg.DATA.RESIZE_W = 224
cfg.DATA.NUM_FRAMES = 25
cfg.DATA.SAMPLING_RATE = 5
cfg.DATA.TRAIN_JITTER_SCALES = [256, 320]
cfg.DATA.TRAIN_CROP_SIZE = 224
cfg.DATA.TEST_CROP_SIZE = 224
cfg.DATA.INPUT_CHANNEL_NUM = [3]
cfg.DATA.PATH_TO_DATA_DIR = "/datasets/clevrer"
cfg.DATA.PATH_PREFIX = "/datasets/clevrer"
cfg.DATA.MAX_TRAIN_LEN = None
cfg.DATA.MAX_VAL_LEN = None
cfg.SOLVER.BASE_LR = 0.001
cfg.SOLVER.LR_POLICY = "cosine"
cfg.SOLVER.COSINE_END_LR = 0.00001
cfg.SOLVER.EPOCH_CYCLE = 10.0
cfg.SOLVER.MAX_EPOCH = 10
cfg.SOLVER.MOMENTUM = 0.9
cfg.SOLVER.NESTEROV = True
cfg.SOLVER.WEIGHT_DECAY = 0.00001
cfg.SOLVER.WARMUP_EPOCHS = 0.0
cfg.SOLVER.WARMUP_START_LR = 0.01
cfg.SOLVER.OPTIMIZING_METHOD = "sgd"
cfg.MODEL.ARCH = "CNN_SEP_LSTM"
cfg.MODEL.MODEL_NAME = "CNN_SEP_LSTM"
cfg.NUM_GPUS = 1
cfg.LOG_PERIOD = 100
cfg.OUTPUT_DIR = "./"
cfg.RNG_SEED = 42
cfg.WORD_EMB.USE_PRETRAINED_EMB = True
cfg.WORD_EMB.TRAINABLE = True
cfg.WORD_EMB.GLOVE_PATH = '/datasets/word_embs/glove.6B/glove.6B.50d.txt'
cfg.WORD_EMB.EMB_DIM = 50
cfg.CLEVRERMAIN.LSTM_HID_DIM = 256
cfg.CLEVRERMAIN.T_DROPOUT = 0.1
return cfg
def pre_proc_config(cfg: CN, dct: Dict = None):
"""
Add any pre processing based on cfg
"""
def upd_sub_mdl(
cfg: CN,
sub_mdl_default_cfg: CN,
sub_mdl_name_key: str,
sub_mdl_file_key: str,
sub_mdl_mapper: Dict,
new_dct: Dict,
):
if new_dct is not None and sub_mdl_name_key in new_dct:
sub_mdl_name = new_dct[sub_mdl_name_key]
else:
sub_mdl_name = CfgProcessor.get_val_from_cfg(
cfg, sub_mdl_name_key)
assert sub_mdl_name in sub_mdl_mapper
sub_mdl_file = sub_mdl_mapper[sub_mdl_name]
assert Path(sub_mdl_file).exists()
CfgProcessor.update_one_full_key(cfg,
{sub_mdl_file_key: sub_mdl_file},
full_key=sub_mdl_file_key)
sub_mdl_default_cfg.merge_from_file(sub_mdl_file)
sub_mdl_cfg = yaml.safe_load(sub_mdl_default_cfg.dump())
sub_mdl_cfg_dct_keep = {k: v for k, v in sub_mdl_cfg.items()}
return CN(sub_mdl_cfg_dct_keep)
sf_mdl_cfg_default = get_cfg()
cfg.sf_mdl = upd_sub_mdl(
cfg,
sf_mdl_cfg_default,
"mdl.sf_mdl_name",
"mdl.sf_mdl_cfg_file",
sf_mdl_to_cfg_fpath_dct,
dct,
)
tx_dec_default = get_default_tx_dec_cfg()
cfg.tx_dec = upd_sub_mdl(
cfg,
tx_dec_default,
"mdl.tx_dec_mdl_name",
"mdl.tx_dec_cfg_file",
tx_to_cfg_fpath_dct,
dct,
)
return cfg
def load_config_from_file(cfg_file=None):
"""
Given the arguemnts, load and initialize the configs.
Args:
args (argument): arguments includes `shard_id`, `num_shards`,
`init_method`, `cfg_file`, and `opts`.
"""
# Setup cfg.
cfg = get_cfg()
# Load config from cfg.
if cfg_file is not None:
cfg.merge_from_file(cfg_file)
# Create the checkpoint dir.
cu.make_checkpoint_dir(cfg.OUTPUT_DIR)
return cfg
def main():
args = parser_args()
print(args)
cfg_file = args.cfg_file
checkpoint_file = args.checkpoint
save_checkpoint_file = args.save
half_flag = args.half
cfg = get_cfg()
cfg.merge_from_file(cfg_file)
cfg.TEST.CHECKPOINT_FILE_PATH = checkpoint_file
print("simplifier model!\n")
with torch.no_grad():
model = build_model(cfg)
model.eval()
cu.load_test_checkpoint(cfg, model)
if half_flag:
model.half()
with open(save_checkpoint_file, 'wb') as file:
torch.save({"model_state": model.state_dict()}, file)
def main():
cfg = get_cfg()
# print(cfg)
split = "train"
drop_last = True
cfg.DATA.PATH_TO_DATA_DIR = ''
cfg.TEST.NUM_ENSEMBLE_VIEWS = 2
dataset = Kinetics(cfg, split)
train_loader = torch.utils.data.DataLoader(
dataset,
batch_size=2,
shuffle=False,
num_workers=cfg.DATA_LOADER.NUM_WORKERS,
pin_memory=cfg.DATA_LOADER.PIN_MEMORY,
drop_last=drop_last,
)
for cur_iter, (inputs, labels, _) in enumerate(train_loader):
print(len(inputs))
print(inputs[0].shape)
print(inputs[1].shape)
print(labels)
def test_train_iter4():
cfg = get_cfg()
cfg.LOG_PERIOD = 2
train_meter = ClevrerTrainMeter(5, cfg)
train_meter.update_stats(top1_err=40.0,
top5_err=20.0,
mc_opt_err=55.0,
mc_q_err=78.0,
loss_des=6.3,
loss_mc=0.2,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
train_meter.update_stats(top1_err=10.0,
top5_err=10.0,
mc_opt_err=10.0,
mc_q_err=10.0,
loss_des=10.0,
loss_mc=10.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
stats = train_meter.log_iter_stats(cur_epoch=1, cur_iter=2)
assert stats is None
train_meter.update_stats(top1_err=20.0,
top5_err=20.0,
mc_opt_err=20.0,
mc_q_err=20.0,
loss_des=20.0,
loss_mc=20.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
stats = train_meter.log_iter_stats(cur_epoch=0, cur_iter=3)
assert stats['top1_err'] == 15.0
assert stats['top5_err'] == 15.0
assert stats['mc_opt_err'] == 15.0
assert stats['mc_q_err'] == 15.0
assert stats['loss_des'] == 15.0
assert stats['loss_mc'] == 15.0
def test_train_iter5():
cfg = get_cfg()
cfg.LOG_PERIOD = 3
train_meter = ClevrerTrainMeter(5, cfg)
train_meter.update_stats(top1_err=30.0,
top5_err=30.0,
mc_opt_err=30.0,
mc_q_err=30.0,
loss_des=30.0,
loss_mc=30.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
train_meter.update_stats(top1_err=10.0,
top5_err=10.0,
mc_opt_err=10.0,
mc_q_err=10.0,
loss_des=10.0,
loss_mc=10.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
stats = train_meter.log_iter_stats(cur_epoch=1, cur_iter=1)
assert stats is None
train_meter.update_stats(top1_err=20.0,
top5_err=20.0,
mc_opt_err=20.0,
mc_q_err=20.0,
loss_des=20.0,
loss_mc=20.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=10)
stats = train_meter.log_iter_stats(cur_epoch=0, cur_iter=5)
assert stats['top1_err'] == 20.0, print(stats['top1_err'])
assert stats['top5_err'] == 20.0
assert stats['mc_opt_err'] == 20.0
assert stats['mc_q_err'] == 20.0
assert stats['loss_des'] == 20.0
assert stats['loss_mc'] == 20.0
def test_train_epoch_only_des():
cfg = get_cfg()
cfg.LOG_PERIOD = 3
train_meter = ClevrerTrainMeter(5, cfg)
train_meter.update_stats(top1_err=30.0,
top5_err=30.0,
mc_opt_err=30.0,
mc_q_err=30.0,
loss_des=30.0,
loss_mc=30.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=0)
train_meter.update_stats(top1_err=10.0,
top5_err=10.0,
mc_opt_err=10.0,
mc_q_err=10.0,
loss_des=10.0,
loss_mc=10.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=0)
train_meter.update_stats(top1_err=20.0,
top5_err=20.0,
mc_opt_err=20.0,
mc_q_err=20.0,
loss_des=20.0,
loss_mc=20.0,
lr=0.1,
mb_size_des=10,
mb_size_mc=0)
stats = train_meter.log_epoch_stats(cur_epoch=1)
assert stats['top1_err'] == 20.0, print(stats['top1_err'])
assert stats['top5_err'] == 20.0
assert not 'mc_opt_err' in stats
assert not 'mc_q_err' in stats
assert stats['loss_des'] == 20.0
assert not 'loss_mc' in stats
)
parser.add_argument(
"--tag",
help="tag",
default="slowfast_r101",
type=str,
)
parser.add_argument(
"--test_tag",
help="test_tag",
default="1",
type=str,
)
args = parser.parse_args()
cfg = get_cfg()
# Merge train configs
cfg.merge_from_file(os.path.join('logdir', args.tag, 'config.yaml'))
# Merge test configs
if args.cfg_file is not None:
cfg.merge_from_file(args.cfg_file)
print("Test using", args.tag)
cfg.OUTPUT_DIR = os.path.join('logdir', args.tag)
cfg.TEST_OUTPUT_DIR = os.path.join(cfg.OUTPUT_DIR, "test_" + args.test_tag)
# Make dir
if not os.path.exists(cfg.TEST_OUTPUT_DIR):
os.makedirs(cfg.TEST_OUTPUT_DIR)
# Save test cfg
copyfile(args.cfg_file,
os.path.join(cfg.TEST_OUTPUT_DIR, 'test_config.yaml'))
def basic_sf_cfg(rel_yml_path):
sf_cfg = get_cfg()
yml_path = (Path(slowfast.__file__)
.parents[1]/f'configs/{rel_yml_path}')
sf_cfg.merge_from_file(yml_path)
return sf_cfg
def test_val_epoch():
cfg = get_cfg()
cfg.LOG_PERIOD = 4
val_meter = ClevrerValMeter(5, cfg)
val_meter.update_stats(top1_err=30.0,
top5_err=30.0,
mc_opt_err=60.0,
mc_q_err=60.0,
loss_des=30.0,
loss_mc=60.0,
mb_size_des=10,
mb_size_mc=5)
val_meter.update_stats(top1_err=10.0,
top5_err=10.0,
mc_opt_err=10.0,
mc_q_err=10.0,
loss_des=10.0,
loss_mc=10.0,
mb_size_des=10,
mb_size_mc=5)
val_meter.update_stats(top1_err=20.0,
top5_err=20.0,
mc_opt_err=20.0,
mc_q_err=20.0,
loss_des=20.0,
loss_mc=20.0,
mb_size_des=10,
mb_size_mc=5)
stats = val_meter.log_epoch_stats(cur_epoch=1)
assert stats['top1_err'] == 20.0, print(stats['top1_err'])
assert stats['top5_err'] == 20.0
assert stats['mc_opt_err'] == 30.0
assert stats['mc_q_err'] == 30.0
assert stats['loss_des'] == 20.0
assert stats['loss_mc'] == 30.0
val_meter.reset()
val_meter.update_stats(top1_err=60.0,
top5_err=30.0,
mc_opt_err=30.0,
mc_q_err=30.0,
loss_des=30.0,
loss_mc=30.0,
mb_size_des=10,
mb_size_mc=5)
val_meter.update_stats(top1_err=10.0,
top5_err=10.0,
mc_opt_err=10.0,
mc_q_err=10.0,
loss_des=10.0,
loss_mc=10.0,
mb_size_des=10,
mb_size_mc=5)
val_meter.update_stats(top1_err=20.0,
top5_err=20.0,
mc_opt_err=20.0,
mc_q_err=20.0,
loss_des=20.0,
loss_mc=20.0,
mb_size_des=10,
mb_size_mc=5)
stats = val_meter.log_epoch_stats(cur_epoch=2)
assert stats['top1_err'] == 30.0, print(stats['top1_err'])
assert stats['top5_err'] == 20.0
assert stats['mc_opt_err'] == 20.0
assert stats['mc_q_err'] == 20.0
assert stats['loss_des'] == 20.0
assert stats['loss_mc'] == 20.0
def load_config(args):
with open(args.cfg_file, "r") as f:
raw_cfg = yaml.load(f, Loader=yaml.CLoader)
cfg = get_cfg()
cfg.update(raw_cfg)
return cfg
