GitLab

* [Dist] Reduce peak memory in DistDGL: avoid validation, release memory once loaded

* remove orig_id from ndata/edata for partition_graph()

* delete orig_id from ndata/edata in dist part pipeline

* reduce dtype size and format before saving graphs

* fix lint

* ETYPE requires to be int32/64 for CSRSortByTag

* fix test failure

* refine

* [Dist] enable iterate multiple dist dataloaders simultaneously

* format file

* add support for any number of dataloaders

* fix lint

* refine code

Co-authored-by: Steve <ubuntu@ip-172-31-34-29.ap-northeast-1.compute.internal>

* first commit

* add test

* fixes

* ah this is how you skip setup

* fix

* ugh

* address comments

* i like black

* [Feature] enable graph partition book support canonical etypes

* fix lint

* fix lint

* add todo

* refine according to review comments

* fix lint

* refine naming

* revert PartitionPolicy __init__

* refine docstring

* fix doc string

* [DistTest] fix incorrect shell if statement

* fix incorrect use of dist.initialize()

* [Dist] enable to specify sort_etype for sample_etype_neighbours

* fix lint

* pass argument instead of env

* fix lint and doc string

* refine args

* remove unnecessary lines

* debug only

* debug add sort time log

* change interface

* fix typo
Co-authored-by: Xin Yao <xiny@nvidia.com>

* [Dist] format dtypes when loading graph in server

* add test

* refine

* add comments

* [Dist] defer to load node/edge feats

* fix lint

* Update python/dgl/distributed/partition.py
Co-authored-by: Minjie Wang <minjie.wang@nyu.edu>

* Update python/dgl/distributed/partition.py
Co-authored-by: Minjie Wang <minjie.wang@nyu.edu>

* fix lint
Co-authored-by: Minjie Wang <minjie.wang@nyu.edu>

* [Dist] set socket as default backend for RPC

* add tests both for socket and tensorpipe

* [ist] enable time out when fetching msg

* fix lint error

* minor refinements

* improve minor log

* fix dist test

* fix timeout issue in tensorpipe

* [Dist][BugFix] enable sampling on bipartite

* add comments for tests

* [Dist] Enable maximum try times for socket backend via DGL_DIST_MAX_TRY_TIMES

* reset env before/after test

* print log for info when trying to connect

* fix

* print log in python instead of cpp

* [Feature] enable socket net_type for rpc

* fix lint

* fix lint

* fix build issue on windows

* fix test failure on windows

* fix test failure

* fix cpp unit test failure

* net_type blocking max_try_times

* fix other comments

* fix lint

* fix comment

* fix lint

* fix cpp

Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>

* [BugFix] pass ntype/etype into partition book when node/edge_split

* fix test failure

* fix test failue on mxnet

* fix test failure

* [Fix] sleep for a while when launching clients which will connect to multiple servers

* pre-allocate more ports

* no multiple partitions on single machine

* initial update

* more

* more

* multi-gpu example

* cluster gcn, finalize homogeneous

* more explanation

* fix

* bunch of fixes

* fix

* RGAT example and more fixes

* shadow-gnn sampler and some changes in unit test

* fix

* wth

* more fixes

* remove shadow+node/edge dataloader tests for possible ux changes

* lints

* add legacy dataloading import just in case

* fix

* update pylint for f-strings

* fix

* lint

* lint

* lint again

* cherry-picking commit fa9f494

* oops

* fix

* add sample_neighbors in dist_graph

* fix

* lint

* fix

* fix

* fix

* fix tutorial

* fix

* fix

* fix

* fix warning

* remove debug

* add get_foo_storage apis

* lint

* migrate to pylint 2.6.0

* fix

* fix?

* ???

* oops

* [Feature] long live server for multiple client groups

* generate globally unique name for DistTensor within DGL automatically

* [Fix] reduce error msg, refine fetch logic of available ports

* un-initialize client before sending shutdown request

* fix import error

* print connect failure log only in debug mode

* enable DMLC_LOG_DEBUG=1 in CI

* [Feature] enable TP::Receiver wait for any numbers of senders

* fix random unit test failure

* avoid endless future wait

* fix unit test failure

* fix seg fault when finalize wait in receiver

* [Feature] refactor sender connect logic and remove unnecessary sleeps in unit tests

* fix lint

* release RPCContext resources before process exits

* [Debug] TPReceiver wait start log

* [Debug] add log in get port

* [Debug] add log

* [ReDebug] revert time sleep in unit tests

* [Debug] remove sleep for test_distri,test_mp

* [debug] add more log

* [debug] add listen_booted_ flag

* [debug] restore commented code for queue

* [debug] sleep more in rpc_client

* restore change in tests

* Revert "restore change in tests"

This reverts commit 41a18926d181ec2517069389bfc41de2cc949280.

* Revert "[debug] sleep more in rpc_client"

This reverts commit a908e758eabca0a6ce62eb2e59baea02a840ac67.

* Revert "[debug] restore commented code for queue"

This reverts commit d3f993b3746e6bb6e2cc2f90204dd7e9461c6301.

* Revert "[debug] add listen_booted_ flag"

This reverts commit 244b2167d94942ff2a0acec8823b974975e52580.

* Revert "[debug] add more log"

This reverts commit 4b78447b0a575a824821dc7e25cca2246e6e30e2.

* Revert "[Debug] remove sleep for test_distri,test_mp"

This reverts commit e1df1aadcc8b1c2a0013ed77322ac391a8807612.

* remove debug code

* revert unnecessary change

* revert unnecessary changes

* always reset RPCContext when get started and reset all data

* remove time.sleep in dist tests

* fix lint

* reset envs before each dist test

* reset env properly

* add time sleep when start each server

* sleep for a while when boot server

* replace wait_thread with callback

* fix lint

* add dglconnect handshake check
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>

* doesn't know whether works

* add change

* fix

* fix

* fix

* remove

* revert

* lint

* lint

* fix

* revert

* lint

* fix

* only build rpc on linux

* lint

* lint

* fix build on windows

* fix windows

* remove old test

* fix cmake

* Revert "remove old test"

This reverts commit f1ea75c777c34cdc1f08c0589676ba6aee1feb29.

* fix windows

* fix

* fix

* fix indent

* fix indent

* address comment

* fix

* fix

* fix

* fix

* fix

* lint

* fix indent

* fix lint

* add introduction

* fix

* lint

* lint

* add more logs

* fix

* update xbyak for C++14 with gcc5

* Remove channels

* fix

* add test script

* fix

* remove unused file

* fix lint

* add timeout

* first commit

* second commit

* spaghetti unit tests

* rewrite test

[Feature] Add a HINT for the per edge type sampler of heterogeneous DistGraph that highlighting the etypes are sorted already. (#3260)

* pass cpp test

* distgraph use sorted edge flag.

* lint

* triger

* update test
Co-authored-by: Ubuntu <ubuntu@ip-172-31-2-66.ec2.internal>

* handle empty frontiers.

* fix lint.

* fix
Co-authored-by: Ubuntu <ubuntu@ip-172-31-2-202.us-west-1.compute.internal>
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>

* Fix dist negative data loader bug

* upd

* Fix
Co-authored-by: Da Zheng <zhengda1936@gmail.com>

* make heterogeneous find_edges

* add distributed EdgeDataLoader.

* fix.

* fix a bug.

* fix bugs.

* add tests on distributed heterogeneous graph sampling.

* fix.
Co-authored-by: Zheng <dzzhen@3c22fba32af5.ant.amazon.com>

* fix.

* fix.

* fix.

* fix.

* Fix test

* Deprecate old DistEmbedding impl, use synchronized embedding impl

* Basic imple of heterogeneous on homogenenous sampling

* make pass

* Pass C++ test

* Add python test code

* lint

* lint

* Add MultiLayerEtypeNeighborSampler

* Add unitest for single machine dataloader

* Add dist dataloader test for edge type sampler

* Fix lint

* fix

* support for per etype sample

* Fix some bug and enable distributed training with per edge sample

* fix

* Now distributed training works

* turn off some mxnet

* turn off mxnet for some dist test

* fix

* upd

* upd according to the comments

* Fix

* Fix test and now distributed works.

* upd

* upd

* Fix

* Fix bug

* remove dead code.

* upd

* Fix

* upd

* Fix
Co-authored-by: Ubuntu <ubuntu@ip-172-31-71-112.ec2.internal>
Co-authored-by: Ubuntu <ubuntu@ip-172-31-2-66.ec2.internal>
Co-authored-by: Da Zheng <zhengda1936@gmail.com>

* support hetero RGCN.

* fix.

* simplify code.

* sample_neighbors return heterograph directly.

* avoid using to_heterogeneous.

* compute canonical etypes in advance.

* fix tests.

* fix.

* fix distributed data loader for heterograph.

* use NodeDataLoader.

* fix bugs in partitioning on heterogeneous graphs.

* fix lint.

* fix tests.

* fix.

* fix.

* fix bugs.

* fix tests.

* fix.

* enable coo for distributed.

* fix.

* fix.

* fix.

* fix.

* fix.
Co-authored-by: Ubuntu <ubuntu@ip-172-31-71-112.ec2.internal>
Co-authored-by: Zheng <dzzhen@3c22fba32af5.ant.amazon.com>

* fix.

* fix.

* fix.

* fix.

* Fix test

* Deprecate old DistEmbedding impl, use synchronized embedding impl

* update doc
Co-authored-by: Ubuntu <ubuntu@ip-172-31-71-112.ec2.internal>
Co-authored-by: Ubuntu <ubuntu@ip-172-31-2-66.ec2.internal>
Co-authored-by: Da Zheng <zhengda1936@gmail.com>
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>

* fix.

* fix.

* fix.

* fix.

* Fix test
Co-authored-by: Ubuntu <ubuntu@ip-172-31-71-112.ec2.internal>
Co-authored-by: Ubuntu <ubuntu@ip-172-31-2-66.ec2.internal>

* fix bugs in partitioning on heterogeneous graphs.

* fix.

* fix.

* fix example.

* fix.

* fix test.

* fix.

* fix.

* fix.

* fix tests.
Co-authored-by: Ubuntu <ubuntu@ip-172-31-71-112.ec2.internal>
Co-authored-by: Zheng <dzzhen@3c22fba32af5.ant.amazon.com>