[Code scan] Close TensorFlow EwaldRecp sessions

[njzjz]

This issue comes from a Codex global scan of deepmodeling/deepmd-kit at commit 73de44b1f94471b2e3bdb6b11f57b34d7bc791bb.

Problem

EwaldRecp creates a dedicated TensorFlow session in its constructor:

deepmd-kit/deepmd/tf/infer/ewald_recp.py

Lines 31 to 50 in 73de44b

	with tf.Graph().as_default() as graph:
	# place holders
	self.t_nloc = tf.placeholder(tf.int32, [1], name="t_nloc")
	self.t_coord = tf.placeholder(
	GLOBAL_TF_FLOAT_PRECISION, [None], name="t_coord"
	)
	self.t_charge = tf.placeholder(
	GLOBAL_TF_FLOAT_PRECISION, [None], name="t_charge"
	)
	self.t_box = tf.placeholder(GLOBAL_TF_FLOAT_PRECISION, [None], name="t_box")
	# output
	self.t_energy, self.t_force, self.t_virial = op_module.ewald_recp(
	self.t_coord,
	self.t_charge,
	self.t_nloc,
	self.t_box,
	ewald_h=self.hh,
	ewald_beta=self.beta,
	)
	self.sess = tf.Session(graph=graph, config=default_tf_session_config)

DipoleChargeModifier creates an EwaldRecp instance and keeps it for modifier evaluation:

deepmd-kit/deepmd/tf/modifier/dipole_charge.py

Lines 75 to 88 in 73de44b

	# the dipole model is loaded with prefix 'dipole_charge'
	self.modifier_prefix = "dipole_charge"
	# init dipole model
	DeepDipole.__init__(
	self, model_name, load_prefix=self.modifier_prefix, default_tf_graph=True
	)
	self.model_name = model_name
	self.model_charge_map = model_charge_map
	self.sys_charge_map = sys_charge_map
	self.sel_type = list(self.get_sel_type())
	# init ewald recp
	self.ewald_h = ewald_h
	self.ewald_beta = ewald_beta
	self.er = EwaldRecp(self.ewald_h, self.ewald_beta)

The scanned modifier and Ewald files do not expose a close method, context-manager path, or destructor that closes EwaldRecp.sess.

Impact

Creating and discarding DipoleChargeModifier / EwaldRecp instances can leak TensorFlow sessions and graph resources. Long-running Python processes that repeatedly construct modifiers can accumulate those resources until process exit.

Suggested fix

Add an explicit close() method to EwaldRecp and have DipoleChargeModifier forward it. Consider context-manager support and a defensive destructor, matching the session lifecycle used elsewhere in the TensorFlow inference stack.