Former-commit-id: 7f87c54d1c4eac4f8d53eee64765f247ab61eb66

Former-commit-id: be7a3cfdc7931d2176ae7e872d8a553627420e81

Former-commit-id: 589d05919b70ea234b73ce3d169007df07050918

Former-commit-id: b03f9869f1305ba08920866b3231b319e7aff093

Former-commit-id: f1a24c8cb002ae895140be7cfb78dbaebf5eeef5

* operator setup

* generate anchors

* prepare for nms

* sort by fg score

* nms

* bbox transform


Former-commit-id: 38cab527e4e87ea25331de1e00ab222e4bae9795

Former-commit-id: 2d4969e86a9b90e1f16ddab71be037da763d7758

Former-commit-id: 8518e1bdb3bf3bf90c5da94a708243cb0808c637

Former-commit-id: 604cab376a780fac2e880f4167570344dbb6d1c3

Former-commit-id: 765c8a31f1da82f73372b88f3a7d0ec6e6afcbe8

Former-commit-id: 9a7a3beec5af6a40e9b029218ad8779707014f30

Former-commit-id: e6bf6a3d3bcc701c96616b8e6e4d3f60611daaff

* add interface and todo

* add comment on output size

* fix minor error

* add channel num

* address review

* InferBlobDescs

* merge master

* remove NeedOutWhenBackward


Former-commit-id: fe11fda46a1d8135019dfe99d05ebccb00e77f0d

Former-commit-id: 1de744e248fc3874f7fb1fc996763c2d302a33a1

* add interface and todo

* add comment on output size

* fix minor error

* add channel num

* address review

* InferBlobDescs


Former-commit-id: 4924e91cf9300521c92c645ca30320d247ea9941

Former-commit-id: 508a3507

Former-commit-id: 72910ebc

* add ctrl edge between copyHd and MdUpdte

* fix bug of add ctrl regst

* hack ctrl regst max regst num

* test undo

* after experiment

* use get task type instead of dynamic cast

* fix for review

* remove hack regst

* init ctrl regst min num

* fix ctrl regst num between copy and mdupdt after improver

* mdupdt actor return ctrl regst k(k = num of piece in batch) one act

* add returned regst num of ctrl regst

* CHECK invariant


Former-commit-id: 339d56ee

Former-commit-id: ad3f5d65

Former-commit-id: 4faea69b

* optimize the add op

* remove useless code


Former-commit-id: 57689ae2

* feat: avoid net contention by adding ctrl edge in ReduceStruct

* refine(task_graph.h/cpp): refine AddCtrlEdgeInReduceStruct()

* fix(graph/task_graph.cpp): fix the bug of machine order

* fix(graph/task_graph.cpp): do not add ctrl edge with reduce scatter

* feat: add ReduceGlobalAddCompActor

* fix: fix the bug of reduce_global_actor/kernel

* chore: remove used vim .swp file

* fix(graph/task_graph.cpp): fix the bug of sorting copycomment when build reduce ctrl edge

* fix(graph/task_graph.h/cpp): add CtrlEdge for ReduceGather

* feat: revert add ctrl edge in reduce struct from this PR

* refactor: rename ReduceGlobalAddCompActor to InputWiseCompActor for scalability

* fix(kernel/reduce_global_add_kernel.cpp): use Memcpy other than Memset for first blob to be added

* refactor(actor/input_wise_compute_actor.*): use HashMap and counter instead of HashSet for processed regst_desc

* refactor: let ReduceGlobalAddCompActor inherit InputWiseCompActor

* feature: add ReduceGatherCompActor that inherits InputWiseCompActor

* fix(reduce_gather_kernel.cpp): add missing break

* refactor: replace regst_desc_id2bn_in_op_ with regst_desc_id2in_bn_id_ in InputWiseCompActor

* fix(reduce_global_add_kernel): remove useless class member parallel_id_

* refactor: make ReduceLocalAdd kernel support inputwise, rename ReduceGlobalAddActor to ReduceAddActor for scalibility


Former-commit-id: 0ffc781c

Former-commit-id: 9a010dae

* naive version of accuracy module

* add top_k_ prefix to accuracy print

* remove magic number


Former-commit-id: 7db6b07b

* speedup clone kernel

* refine

* simplify

* handle with arbitrary number of out_diff


Former-commit-id: 9d0bae73

* let NormalMdUpdtTask use independent stream

* let kMix task share the same stream

* remove UseIndependentStream

* refine


Former-commit-id: cb3f3f37

Former-commit-id: 7ab4cee4

* feat: avoid net contention by adding ctrl edge in ReduceStruct

* refine(task_graph.h/cpp): refine AddCtrlEdgeInReduceStruct()

* fix(graph/task_graph.cpp): fix the bug of machine order

* fix(graph/task_graph.cpp): do not add ctrl edge with reduce scatter

* feat: add ReduceGlobalAddCompActor

* fix: fix the bug of reduce_global_actor/kernel

* chore: remove used vim .swp file

* fix(graph/task_graph.cpp): fix the bug of sorting copycomment when build reduce ctrl edge

* fix(graph/task_graph.h/cpp): add CtrlEdge for ReduceGather

* revert: remove the ReduceGlobalAddCompActor from this PR

* feat: add use_ordered_allreduce_in_mdupdt in OtherConf


Former-commit-id: 6d9663c7

* refactor comm_network

* refine

* AddWorkToStream

* refine

* add CHECK

* refine


Former-commit-id: ea965893

* alexnet, resnet and inception v3 generator

* move net gen apps to tools


Former-commit-id: a68583b3

* bugfix: fix node2component_id bug according to new depth definition

* only take account of ActNode with output


Former-commit-id: 4a0f6219

Former-commit-id: d11e1a04

Former-commit-id: b2d088d4

Former-commit-id: 9a486ad1

* carefully handle depth of act_node

* revert dfs topo


Former-commit-id: 81e29440

* carefully handle depth of act_node

* bfs topo instead of dfs topo


Former-commit-id: d5785541

Former-commit-id: 96ca9d84

* fix bug

* reduce one copy


Former-commit-id: e1620821

* perform rpc call in background thread pool

* comment: stream poller thread is not allowed to perform blocking action


Former-commit-id: 3a85b18a

Former-commit-id: 54347560