* source subset tick

* remove useless header files

* insert DstSubsetTickOp

* remove incorrect CHECK

* TryBindBnWithOneofRegst

* fix typo in task_graph

* rename and refactor DstSubsetTick::InferBlobDescs and SrcSubsetTick::InferBlobDescs

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* Print backtrace when distributed CI failed

* fix path

* typo

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* Add Operator::InferInternalBlobDescs

* XrtLaunchOp::InferOutBlobDescs

* refine

* add weight_decay arguments

* add weight_decay test

* code format

* add weight_decay note

* Packed all separated mem block for allocate memory speed up

* only packed separated mem block used by same thread.

* refine implement; add thrd_id to mem_block

* refine name

* improve implement

* Remove OpContext

* fix

* Add Operator::InferInplaceObn2IbnIf

* remove useless header

* make InferInplaceObn2Ibn protected

* add pad2d ops and kernels

* fix bug

* add python api and unittest

* reformat, change padding_data_type seq, change dim2vector method usage

* fix typo

* delete ShapeViewToDimVector

* rerun make of_format

* save model and load model demo

* fix

* tensor.proto and copy signature

* pass test

* add load_saved_model function for InferenceSession

* wait_for_all_jobs_finished

* test_alexnet_save_and_load

* support change batch_size

* support batch axis

* add ci test

* revert job_build_and_infer_ctx api

* simplify test script

* following update

* improve search function of InferenceSession

* fix break update

* add cv2 to dev-requirements

* Update Dockerfile

* rm 3.5

* fix

* quick workaround

* speed up bazel

* port changes

* revert workaround

* rm batch_axis in JobInputDef and JobOutputDef

* rm export for ImageNetRecordDataset

* refine Complete api for GraphBuilder and SignatureBuilder

* refine check op is moirrored

* fix by review comment ci test

* InferenceSession is not responsible to destroy env

Co-authored-by: Shenghang Tsai <jackalcooper@gmail.com>
Co-authored-by: Tsai <caishenghang@oneflow.org>
Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* Remove useless code

* Remove OpGraph::MakeGetterBlobDesc4ModelLbi

* Remove ExecNode::fw_node_

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* port with cuda

* refactor generated_compile_flags with cuda

* use_cxx11_abi

* refactor

* larger tol

Co-authored-by: Tsai <caishenghang@oneflow.org>
Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* refactor boxing_sub_task_builder

* refine

* refine

* refine

* refine

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* Interface alignment

* format code

* update padding param

* update test case

* format

* reduction

* update test case

* format

* fix slice_boxing s2s multi-machine when empty intersection

* refine

Co-authored-by: Juncheng <liujuncheng1022@gmail.com>
Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* mig parallel_conf_util

* mig BuildInitialScope BuildScopeWithNewParallelDesc BuildScopeWithNewParallelConf

* add test of GetDeviceTagAndMachineDeviceIds

* fix BuildScopeWithNewParallelDesc input type error

* use TRY

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Signed-off-by: daquexian <daquexian566@gmail.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* manylinux docker use pip_args

* optional bazel url

* move args

* fix repo url

* reorder cmd

* fix github case

* update manylinux sha

* http proxy lower case

* rm err msg

* mv msg

* fix case

* add exit 1

* disable centos-sclo-rh

* centos-sclo-rh skip_if_unavailable

* MANYLINUX_SHA

* Update Dockerfile

* Update Dockerfile

* refine

* use ali

* port more changes

* use oneflow url

* it works

* add rsync

* reorder

* refine

* refine

* refine

* use mirrot install cpython

* larger tol

Co-authored-by: Tsai <caishenghang@oneflow.org>

* replace python id_generator with c++

* mig InstructionsBuilder's data to c++

* mig_instruction_build_symbol_api

* mig new rel api

* mig some api

* mig some other api

* mig the rest api

* rename release_object to object_releaser

* move InstructionsBuilder to deprecated MODULE

* mig PackPhysicalBlobsToLogicalBlob

* mig UnpackLogicalBlobToPhysicalBlobs

* move head file

* use Maybe

* remove CHECK_JUST

* use CHECK_EQ_OR_RETURN

* use wraper

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: Tsai <caishenghang@oneflow.org>

* replace python id_generator with c++

* mig InstructionsBuilder's data to c++

* mig_instruction_build_symbol_api

* rename release_object to object_releaser

* move InstructionsBuilder to deprecated MODULE

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* replace python id_generator with c++

* mig InstructionsBuilder's data to c++

* rename release_object to object_releaser

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* parallel_cast user op

* replace parallel_cast in PruneParallelCastOpsPass

* replace parallel_cast in IndexedSlicesOptimizerRewritePass

* replace parallel_cast in DoParallelCastBeforeWideningTypeCast

* replace parallel_cast in autograd

* rm old parallel_cast op

* rm special check of parallel_cast in auto mixed precision

* fix typo

* update PruneParallelCastOpsPass

* improve code

* use string attr to indicate sbp

* make common sbp_parallel util function

* modify by review

Co-authored-by: guo ran <360112263@qq.com>
Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* mig EagerPhysicalBlobHeader

* slove dtype

* mig EagerPhysicalBlob partical

* mig EagerBlobTrait

* fix EagerBlobTrait shape property

* add CHECK

* slove numpy

* enroll blob_trait

* replace EagerPhysicalBlob with oneflow_api.EagerPhysicalBlob

* replace LazyBlob with oneflow_api.LazyBlob

* fix a SyntaxError

* mig eager_blob

* replace EagerBlob with oneflow_api.EagerBlob

* move parallel_size to c++

* Adjust parameter order

* rename fun

* adjust condition

* del useless fun

* refactor_string_symbol

* run make of_format

* rename blob_type to blob_class

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* mig EagerPhysicalBlobHeader

* slove dtype

* mig EagerPhysicalBlob partical

* mig EagerBlobTrait

* fix EagerBlobTrait shape property

* add CHECK

* slove numpy

* enroll blob_trait

* replace EagerPhysicalBlob with oneflow_api.EagerPhysicalBlob

* replace LazyBlob with oneflow_api.LazyBlob

* fix a SyntaxError

* mig eager_blob

* replace EagerBlob with oneflow_api.EagerBlob

* move parallel_size to c++

* Adjust parameter order

* rename fun

* adjust condition

* del useless fun

* rename blob_type to blob_class

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>

* mig EagerPhysicalBlobHeader

* slove dtype

* mig EagerPhysicalBlob partical

* mig EagerBlobTrait

* fix EagerBlobTrait shape property

* add CHECK

* GetPhysicalOpArgBlobAttr

* cfg hash

* fix bug

* cfg::SbpParallel typed OpArgParallelAttribute.sbp_parallel

* cfg::OptMirroredParallel typed OpArgParallelAttribute.opt_mirrored_parallel

* rename PyLazyConsistentBlob and PyLazyMirroredBlob

* fix EagerConsistentBlob bug of property parallel_size

* use static_cast

* del redefine hash

* mig_op_arg_para_attr

* mig DumpToInterfaceBlobConf

* fix bug

* fix inter_face_blob_conf.proto

* replace None with oneflow_api.INVALID_BATCH_AXIS

* migrate python OpNodeSignatureSymbol to c++ version

* fix CONFLICT

* mig DumpToOpNodeSignature

* mig op_arg_util.py completely

* mig python BlobObject

* add del function

* add cortol in del

* fix BlobObject construct

* add const

* fix eager multiple gpu bug

* add shutdown() switch

* add shutdown in vm_util

* fix vm util bug

* fix vm util bug

* fix release function

* fix _DeleteObject and _TryClearObject

* bind is_shutting_down

* mig blob_register

* use void

* del useless line

* add release

* del useless export

Co-authored-by: lixinqi <lixinqi0703106@163.com>
Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>