python/hlo/optimize.py (3 lines): - line 428: # TODO: this will be unnecessary once BatchHloInstructionPool is fully populated - line 685: # TODO: debug this - line 708: # TODO: check collapsed_slice_dims runtime/init.cc (1 line): - line 110: // TODO: fix once we figure out why TF_InitGraph is always called twice grappler/convert/convert_graph.cc (1 line): - line 701: // TODO: determine if we need grappler::TopologicalSort python/graph_def_util.py (1 line): - line 395: TODO: Some passes introduced recently can change subgraph input/output runtime/direct/placer.cc (1 line): - line 149: // TODO: implement NEURONCORE_GROUP_SIZES + automatic data parallel python/_trace.py (1 line): - line 481: # TODO: remove this hack once https://github.com/tensorflow/tensorflow/blob/v2.3.1/tensorflow/python/eager/wrap_function.py#L377 is fixed runtime/direct/adaptor.cc (1 line): - line 169: constexpr int NRT_HOST_NEURON_CORE_ID = -1; // TODO: get from nrt.h