tzrec/protos/feature.proto (10 lines):
	- line 102: // TODO: constrains
	- line 147: // TODO: constrains
	- line 193: // TODO: constrains
	- line 270: // TODO: optional string kv_separator = 13 [default = ":"];
	- line 271: // TODO: constrains
	- line 489: // TODO: constrains
	- line 560: // TODO: constrains
	- line 612: // TODO: constrains
	- line 653: // TODO: constrains
	- line 722: // TODO: constrains


tzrec/acc/export_utils.py (3 lines):
	- line 84: # TODO Remove this allowlist.
	- line 95: # TODO (tmanlaibaatar)
	- line 114: # TODO (tmanlaibaatar) more proper way is needed here


tzrec/datasets/data_parser.py (3 lines):
	- line 450: # TODO: remove to_float when segment_reduce support int values
	- line 575: # TODO: remove to_float when segment_reduce support int values
	- line 675: # TODO: remove to_float when segment_reduce support int values


tzrec/utils/dist_util.py (2 lines):
	- line 241: # TODO: resolve incompatibility with different contexts
	- line 265: TODO: use torchrec DistributedModelParallel when torchrec fix it.


tzrec/features/match_feature.py (1 line):
	- line 209: # TODO: value_type -> int64


tzrec/ops/triton/triton_hstu_attention.py (1 line):
	- line 1472: # TODO (linjianma): avoid hardcoding the value.


tzrec/features/sequence_feature.py (1 line):
	- line 409: # TODO: support dense embedding for sequence raw feature.


tzrec/datasets/sampler.py (1 line):
	- line 778: # TODO: only use one conditional smapler