Summary: 21 instances, 21 unique

Text	Count
# TODO lofe: should also check for ATTR_RANGE	1
# TODO: file name may get changed later	1
# TODO: investigate the cache effect	1
# FIXME: can we make it common	1
# TODO lofe: support kwargs too.	1
# TODO: check the correctness of root, should be between 0 to [world_size -1]	1
# TODO: Error handle	1
# FIXME: alwasy perfom blocking comms; may study non-blocking in the future	1
self.allreduce_qcomm = 32  # TODO: set it as the bitwidth for now until the quantization kernels be supported	1
# TODO lofe: support kwargs too.	1
# TODO: collect perf. from all ranks to rank 0 and detect any imbalanced perf?	1
temp.append(args.arch_sparse_feature_size)  # PENDING/TODO: Update it based on trainer/feeds model.	1
### TODO: add these to class variables?	1
# TODO: allow user to set specific size	1
# TODO: Is this the best way to exit?	1
# FIXME: can we refere to extend_distributed.ExtendProcessGroup.alltoallv?	1
# TODO:	1
# FIXME:  0 is a common case, need this info from trace for more accurate replay	1
# TODO: this is a temporary workaround; need to unify the type of commsParams in comms and dlrm	1
# TODO: only support pytorch-dist as the nw-stack now	1
# FIXME: assuming it's always sum for reduce/allreduce operations	1