[BUG] PT: DpLoaderSet seems setting a fixed seed?

[njzjz]

I notice that in the DpLoaderSet class, there is a seed argument, with the default value of 10.

deepmd-kit/deepmd/pt/utils/dataloader.py

Lines 75 to 84 in 70aa2dd

	def __init__(
	self,
	systems,
	batch_size,
	type_map,
	seed=10,
	shuffle=True,
	):
	setup_seed(seed)
	if isinstance(systems, str):

When constructing DpLoaderSet, no seed is passed, so it seems that it always sets a seed of 10, which does not make sense.

deepmd-kit/deepmd/pt/entrypoints/main.py

Lines 137 to 151 in 70aa2dd

	validation_data_single = (
	DpLoaderSet(
	validation_systems,
	validation_dataset_params["batch_size"],
	model_params_single["type_map"],
	)
	if validation_systems
	else None
	)
	train_data_single = DpLoaderSet(
	training_systems,
	training_dataset_params["batch_size"],
	model_params_single["type_map"],
	)
	return (