* CUDA support max/min reducer on forward pass
* docstring
* concised UpdateGradMinMax_hetero
* reorganized UpdateGradMinMax_hetero
* CUDA kernels for max/min reducer
* variable name
* lint check
* changed CUDA 2D thread mapping to 1D
* removed legacy cusparse for min/max reducer
* git CI issue
* restarting git CI
* adding namespace std
Co-authored-by: Israt Nisa <nisisrat@amazon.com>
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>
* Feat: support API "randomwalk_topk" in library
* Feat: use the new API "randomwalk_topk" for PinSAGESampler
* Minor
* Minor
* Refactor: modified codes as checker required
* Minor
* Minor
* Minor
* Minor
* Fix: checking errors in RandomWalkTopk
* Refactor: modified the docstring for randomwalk_topk
* change randomwalk_topk to internal
* fix
* rename
* Minor for pinsage.py
* Feat: support randomwalk and SelectPinSageNeighbors on GPU
Port RandomWalk algorithm on GPU,
and port SelectPinSageNeighbors on GPU.
* Feat: support GPU on python APIs
* Feat: remove perf print information in FrequenchHashmap
* Fix: modified the code format
Modified the code format as task_lint.sh suggested
* Feat: let test script support PinSAGESampler on GPU
Let test script support PinSAGESampler on GPU,
minor of "restart_prob".
* Minor
* Minor
* Minor
* Refactor: use the atomic operations from the array module
* Minor: change the long lines
* Refactor: modified the get_node_types for gpu
* Feat: update the contributor date
* Perf: remove unnecessary stream sync
* Feat: support other random walk
But the non-uniform choice is still not supported.
* Fix: add CUDA switch for random walk
Co-authored-by: Quan Gan <coin2028@hotmail.com>
* min/max support for forward CPU heterograph
* Added etype with each argU values
* scatter_add needs fix
* added scatter_add_hetero. Grads dont match for max reducer
* storing ntype in argX
* fixing scatter_add_hetero
* hetero matches with torch's scatter add
* works copy_e forward+cpu
* added backward for copy_rhs
* Computes gradient for all node types in one kernel
* bug fix
* unnitest for max/min on CPU
* renamed scatter_add_hetero to update_grad_minmax_hetero
* lint check and comment out cuda call for max. Code is for CPU only
* lint check
* replace inf with zero
* minor
* lint check
* removed LIBXSMM code from hetro code
* fixing backward operator of UpdateGradMinMaxHetero
* removed backward from update_grad_minmax_hetero
* docstring
* improved docstring and coding style
* Added pass by pointer for output
* typos and pass by references
* Support for copy_rhs
* Added header <string>
* fix bug in copy_u_max
* Added comments and dimension check of all etypes
* skip mxnet check
* pass by pointer output arrays
* updated docstring
Co-authored-by: Israt Nisa <nisisrat@amazon.com>
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>
* [Feature] enable async transfer in NodeDataLoader for homograph
* fix lint issues
* fix device choose when creating stream
* fix test on cpu only machine
* fix pin_memory config
* support homo only
* avoid creating stream in each step and sync via event
* fix lint
* enable graph copy on non-default stream
* fix lint
* refine arg description
* fix conflicts
* Based on issue #3436. Improving _SegmentCopyKernel s GPU utilization by switching to nonzero based thread assignment
* fixing lint issues
* Update cub for cuda 11.5 compatibility (#3468)
* fixing type mismatch
* tx guaranteed to be smaller than nnz. Hence removing last check
* minor: updating comment
* adding three unit tests for csr slice method to cover some corner cases
Co-authored-by: Abdurrahman Yasar <ayasar@nvidia.com>
Co-authored-by: nv-dlasalle <63612878+nv-dlasalle@users.noreply.github.com>
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>
* relabel gpu
* unittest for ralebl_ on the GPU
* finish Relabel_ for the GPU
* copyright
* re-enable the unittest for edge_subgrah on the GPU
* fix unittest for tensorflow
* use a fixed number of threads
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>
Co-authored-by: nv-dlasalle <63612878+nv-dlasalle@users.noreply.github.com>
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>
* gpu compact graph template
* cuda compact graph draft
* fix typo
* compact graphs
* pass unit test but fail in training
* example using EdgeDataLoader on the GPU
* refactor cuda_compact_graph and cuda_to_block
* update training scripts
* fix linting
* fix linting
* fix exclude_edges for the GPU
* add --data-cpu & fix copyright
* [Feature] enable create/set/free cuda stream for internal use
* add unit test
* fix unit test failure on mxnet and tf
* refactor stream wrapper
* fix lint error
* fix lint error
* [Feature] Exclude edges in sample_neighbors
Extending sample_neighbors and sample_frontier
API to support exclude_edges parameter.
exclude_edges support tensor and dict data
Feature enable excluding certain edges
during neighborhood sampling
Exclude_edges contains EID's of edges
which will be excluded
during neighbor picking for seed nodes.
Added test case for heterograph and homograph
RFC issue id: 2944
* compatibility
* fix
* fix
Co-authored-by: Quan Gan <coin2028@hotmail.com>
* [Performance] improve coo2csr space complexity when row is not sorted
* [Perf] replace std::vector<> by NDArray
* keep both impl of unsorted coo to csr and choose according to graph density dynamically
* refine criteria to choose btw Unsorted algos
Co-authored-by: Ubuntu <ubuntu@ip-172-31-34-27.us-west-2.compute.internal>
* rgcn with new heterograph API
* apply_edge() forward for multi relation
* undoing changes from rgcn-hetero
* backward apply_edge(copy_u) added
* unittest for apply_edge(copy_e)
* Compatible with new PRs
* resolving conflict with master
* Bringing back change after resolving conflict
* minor
* minor
Co-authored-by: Israt Nisa <nisisrat@amazon.com>
* Make column double indices lazy
* Copy indices to proper contexts
* Fix initialization
* Add unit test
* Fix unit test for tensorflow
* Remove unused member
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>
* Implement range based NDArrayPartition
* Finish implement range based partition support
* Add unit test
* Fix whitepace
* Add Kernel suffix
* Fix argument passing
* Add doxygen docs and improve variable naming
* Add unit test
* Add function for converting a partition book
* Add example to partition_op docs
* Fix dtype conversion for mxnet and tensorflow
* Add pytorch-direct version
* remove
* Add multi-gpu unified tensor test for pytorch
* relocate verification step to each process
* reduce number of workers
* add parameter
Co-authored-by: shhssdm <shhssdm@gmail.com>
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>
Israt Nisa