[Code scan] Fix normal-distribution random disturbances

[njzjz]

This issue is from a Codex global scan of the repository.

create_random_disturb.py accepts -dstyle normal, but that path raises a NumPy argument error.

Evidence:

dpgen/dpgen/data/tools/create_random_disturb.py

Lines 33 to 42 in 7af5246

	def gen_random_disturb(dmax, a, b, dstyle="uniform"):
	d0 = np.random.rand(3) * (b - a) + a
	dnorm = np.linalg.norm(d0)
	if dstyle == "normal":
	dmax = np.random.standard_normal(0, 0.5) * dmax
	elif dstyle == "constant":
	pass
	else:
	# use if we just wanna a disturb in a range of [0, dmax),
	dmax = np.random.random() * dmax

dpgen/dpgen/data/tools/create_random_disturb.py

Lines 291 to 317 in 7af5246

	parser = argparse.ArgumentParser(
	description="Script to generate random disturb configurations"
	)
	parser.add_argument("fin", type=str, help="input file name")
	parser.add_argument("nfile", type=int, help="number of files to be created")
	parser.add_argument("dmax", type=float, help="dmax")
	parser.add_argument(
	"-etmax",
	type=float,
	default=0,
	help="etmax for random strain tensor generations",
	)
	parser.add_argument(
	"-diag",
	type=int,
	default=0,
	help="only diagonal elements of strain tensors are randomized?",
	)
	parser.add_argument("-ofmt", type=str, default="lmp", help="output fileformat")
	parser.add_argument(
	"-dstyle",
	type=str,
	default="uniform",
	help="random distribution style [uniform?]",
	)
	parser.add_argument(
	"-wd",

The parser exposes -dstyle and gen_random_disturb() checks for dstyle == "normal", but calls np.random.standard_normal(0, 0.5). NumPy's standard_normal accepts only the optional size positional argument, so this raises TypeError instead of generating a normally distributed disturbance.

Expected behavior: use an API such as np.random.normal(loc=0, scale=0.5) or otherwise implement the intended normal-distribution scaling.