* 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>
* add output device for dataloading
* Update dataloader
* Get sampler device from dataloader
* Fix line length
* Update examples
* Fix to_block GPU for empty relation types
* Handle the case where the DistGraph has None for the underlying graph
Co-authored-by: Da Zheng <zhengda1936@gmail.com>
* Split from NCCL PR
* Fix type in comment
* Expand documentation for sparse_all_to_all_push
* Restore previous behavior in example
* Re-work optimizer to use NCCL based on gradient location
* Allow for running with embedding on CPU but using NCCL for gradient exchange
* Optimize single partition case
* Fix pylint errors
* Add missing include
* fix gradient indexing
* Fix line continuation
* Migrate 'first_step'
* Skip tests without enough GPUs to run NCCL
* Improve empty tensor handling for pytorch 1.5
* Fix indentation
* Allow multiple NCCL communicator to coexist
* Improve handling of empty message
* Update python/dgl/nn/pytorch/sparse_emb.py
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>
* Update python/dgl/nn/pytorch/sparse_emb.py
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>
* Keepy empty tensor dimensionaless
* th.empty -> th.tensor
* Preserve shape for empty non-zero dimension tensors
* Use shared state, when embedding is shared
* Add support for gathering an embedding
* Fix typo
* Fix more typos
* Fix backend call
* Use NodeDataLoader to take advantage of ddp
* Update training script to share memory
* Only squeeze last dimension
* Better handle empty message
* Keep embedding on the target device GPU if dgl_sparse if false in RGCN example
* Fix typo in comment
* Add asserts
* Improve documentation in example
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>
xiang song(charlie.song)