tf: handle type embedding in DescrptSeA.serialize

[github-actions]

not sure how to handle type embedding - type embedding is not a model parameter,
but instead a part of the input data. Maybe the interface should be refactored...

Line: 1429

deepmd-kit//deepmd/tf/descriptor/se_a.py

Lines 1426 to 1436 in 76371ce

	raise NotImplementedError("spin is unsupported")
	assert self.davg is not None
	assert self.dstd is not None
	# TODO: tf: handle type embedding in DescrptSeA.serialize
	# not sure how to handle type embedding - type embedding is not a model parameter,
	# but instead a part of the input data. Maybe the interface should be refactored...
	return {
	"@class": "Descriptor",
	"type": "se_e2_a",
	"@version": 1,

[njzjz]

@wanghan-iapcm @iProzd We don't plan to support se_e2_a + type_embedding in dpmodel and pt. Do we plan to support self-serialization and deserialization in TF?

[wanghan-iapcm]

I would say no, because se_e2_a + type_embedding is actually implemented by dpa-1 without attention layer.

[github-actions]

Closed in commit 73dab63