No module named 'scatter_cuda' #50
Comments
|
Can you post the log of the install command: |
|
ok, here is goes: (pytorch) alekseic@DL4:~/pytorch_scatter$ nvcc --version (pytorch) alekseic@DL4:~/pytorch_scatter$ rm -rf build && python setup.py install Installed /home/DML/anaconda/envs/pytorch/lib/python3.6/site-packages/torch_scatter-1.0.4-py3.6-linux-x86_64.egg |
|
Matthias, I managed to install everything. It seems that I think that the problem was because when I first installed pytorch_geometric, I was also installing CUDA drivers, and there was some mismatch, and so the 'cuda' versions were not created. Afterwards, when I rebooted the system, CUDA became available, but running You may disregard all this rationalization, I'm a total newbie in python packages installations. |
|
Glad that it‘s now working. It is kinda hard for me to resolve those issues, so thank you for fixing them by yourself. |
* added HGT DBLP example * typo * Merge PyG master (#52) * Adding the Facebok Page-Page dataset * type hints * documentation CI * py 3.8 * fix links * fix links * fail on warning * fail on warning * fix doc Co-authored-by: benedekrozemberczki <benedek.rozemberczki@gmail.com> * revert * Fix Documentation Rendering (#51) * fix doc rendering * fix linting * retrigger checks * remove pytorch 1.7.0 legacy code (#50) * Fix `copy.deepcopy` within lazy `nn.dense.Linear` (#44) * fix deepcopy within lazy Linear * fix merge * assert exception * example to doc * resolve conflict * resolve conflict * Add Figure and Equation to `to_hetero` docstring (#60) * add tex * add svg + docstring * typo * added equation * Message Passing Hooks (#53) * add hooks * docstring * add docstring * allow modification of inputs/output * add test for modifying output * add additional asserts for modifying output test * Rename `HeteroData.get_edges` and `HeteroData.get_nodes` (#58) * rename to_edges and to_nodes * typo * `HeteroConv` (#64) * clean heteroconv * init * init * clean up Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * fix documentation * bipartite function * fix test CI * remove pillow version * clean up for merge * Merge PyG master (#69) * renaming: PointConv to PointNetConv * Fix a broken link in datasets/gdelt.py (#2800) * fix test * re-add batching of strings * add quick start table * gnn cheatsheet * remove pillow version Co-authored-by: Dongkwan Kim <todoaskit@gmail.com> * re-merge * add lazy column to GNN cheatsheet (#70) * `to_hetero_with_bases(model)` (#63) * update * fix linting * basisconv * add ValueError * to_hetero_with_bases impl done * add test * add comments * add comments * docstring * typo * update figure * svg * typo * add test * update * add rgcn equality test * typos * update * typos * update figures * generate new svgs * fix assignment * rename * delete sorted edge types * rename * add legend * fix typo * Test: Check equal outputs of `to_hetero` and `RGCNConv` (#59) * check equal output * add sparsetensor test * check equal output * add sparsetensor test * rename * linting * add missing import * `HeteroData` support for `T.NormalizeFeatures` (#56) * normalize features * allow normalization of any feature * in-place div * normalize features * allow normalization of any feature * in-place div * fix test * no need to re-assign * `HeteroData` support for `T.AddSelfLoops` (#54) * hetero support for AddSelfLoops * check for edge_index attribute * f-string * retrigger checks * revert bipartite changes * hetero support for AddSelfLoops * check for edge_index attribute * f-string * retrigger checks * revert bipartite changes * merge master * merge master * `HeteroData` support for `T.ToSparseTensor` (#55) * hetero support for ToSparseTensor * add test * customize the attribute of SparseTensor.value * rework sort_edge_index * hetero support for ToSparseTensor * add test * customize the attribute of SparseTensor.value * rework sort_edge_index * linting * `HeteroData` support for `T.ToUndirected` (#57) * to_undirected * revert bipartite changes * coalesce + undirected enhancement * merge master * revert bipartite changes * coalesce + undirected enhancement * merge master * clean up * new default relation type * fix tests * resolve merge conflicts * resolve merge conflicts 2 * resolve merge conflicts 3 * Merge PyG master (#74) * renaming: PointConv to PointNetConv * Fix a broken link in datasets/gdelt.py (#2800) * fix test * re-add batching of strings * add quick start table * gnn cheatsheet * remove pillow version * clean up doc for to_dense_batch * clean up * add legend to cheatsheet * Improve terminology (#2837) I think the previous version of the document uses the term 'symmetric' incorrectly. A symmetric matrix is a square matrix that is is equal to its transpose (https://en.wikipedia.org/wiki/Symmetric_matrix). However, the text is only talking about the shape of the matrix, not its content. Hence, 'square (matrix)' would be the correct term to use. * Add batch_size input to to_dense_batch (#2838) * Add batch_size input to to_dense_batch * to_dense_batch fix typo in batch_size param use * add typehints Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * typo * Added return_attention_weights to TransformerConv. (#2807) * added return_weights functionality to tranformer * added return attn weights tests * flake8 * added typehints Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * MD17 (#2843) * Added MD17 dataset * Updated Documentation * Added link to sGDML website in doc * fixed typos in doc and made train variable description clearer * clean up * fix linting * fix doc warning Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * update doc * remove forward doc * add static graph support info to cheatsheet * fix num_nodes in case edge_index is empty * fix math formula * faster GDC import * lazy import * lazy import for datasets * lazy import for nn * Sequential jittable + traceable * typo * typo * update doc Co-authored-by: Dongkwan Kim <todoaskit@gmail.com> Co-authored-by: Markus <markus.zopf@outlook.com> Co-authored-by: Jimmie <jimmiebtlr@gmail.com> Co-authored-by: Jinu Sunil <jinu.sunil@gmail.com> Co-authored-by: Moritz R Schäfer <moritz.schaefer@protonmail.com> * re-add * GraphGym cleaned version (#82) * GraphGym cleaned version * remove deepsnap dependency * fix lint errors, part 1 * fix all lint errors * fix all lint errors * fix all lint errors * apply yapf * Update .gitignore * Integrate GraphGym into PyG (#85) * GraphGym cleaned version * remove deepsnap dependency * fix lint errors, part 1 * fix all lint errors * fix all lint errors * fix all lint errors * apply yapf * Integrate graphgym into pyg, keep user API in project root * fix merge conflict * fix lint errors * Make optional dependencies * merge LICENSE from GraphGym * add import * clean up LICENSE * fix import * resolve merge conflicts * resolve merge conflicts 2 * Merge PyG master (#87) * renaming: PointConv to PointNetConv * Fix a broken link in datasets/gdelt.py (#2800) * fix test * re-add batching of strings * add quick start table * gnn cheatsheet * remove pillow version * clean up doc for to_dense_batch * clean up * add legend to cheatsheet * Improve terminology (#2837) I think the previous version of the document uses the term 'symmetric' incorrectly. A symmetric matrix is a square matrix that is is equal to its transpose (https://en.wikipedia.org/wiki/Symmetric_matrix). However, the text is only talking about the shape of the matrix, not its content. Hence, 'square (matrix)' would be the correct term to use. * Add batch_size input to to_dense_batch (#2838) * Add batch_size input to to_dense_batch * to_dense_batch fix typo in batch_size param use * add typehints Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * typo * Added return_attention_weights to TransformerConv. (#2807) * added return_weights functionality to tranformer * added return attn weights tests * flake8 * added typehints Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * MD17 (#2843) * Added MD17 dataset * Updated Documentation * Added link to sGDML website in doc * fixed typos in doc and made train variable description clearer * clean up * fix linting * fix doc warning Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * update doc * remove forward doc * add static graph support info to cheatsheet * fix num_nodes in case edge_index is empty * fix math formula * faster GDC import * lazy import * lazy import for datasets * lazy import for nn * Sequential jittable + traceable * typo * typo * update doc * Simple models (#2869) * Inclusion of new backbone models * Eliminating head from asap.py * small correction * Create test_gcn.py * Update __init__.py * Update test_gcn.py * Left only the convolutional simple models * Tests included * update * clean up * clean up v2 * fix activation Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * Example for MemPooling. (#2729) * example for mem pooling * backprop on kl loss is done at the end of an epoch. Keys in memory layers are trained only on kl loss. * added learning rate decay. Using PROTIENS_full * flake8 * reduced lr. increased weight decay * changed download location * added comments * clean up Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * typos * fix removeisolatednodes transform in case 'data.num_nodes' is present * fix XConv with dilation > 1 * fix XConv with dilation > 1 * rgcn link prediction (#2734) * implemented LinkPrediction dataset for loading FB15k237 * implemented evaluation for relational link prediction * implemented R-GCNConf link prediction example * fixed bug: wrong initial objects in negative_sampling * changed file downloader urllib.request.urlretrieve to pytorch.data.download_url; renamed LinkPrediction class to RelationalLinkPredictionDataset * update dataset * update example script * rename Co-authored-by: Moritz <moritzblum> Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * fix gnnexplainer draw kwargs * remove python-louvain dependency * allow customization of output in MP jit mode * fix test for py3.6 * changed normalisation to same norm from instance norm to be robust to small var (#2917) * add CITATION.cff * format * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * [ci skip] * add basetransform ABC (#2924) * clean up BaseTransform * clean up GATConv and add comments * add max_num_neighbors as an additional argument * fix jit GATConv on PyTorch 1.8.0 * fix doc * fix gnn explainer with existing self-loops * Rgcn link pred fix (#2946) * added regularization, removed typo in test * clean up Co-authored-by: Moritz <moritzblum> Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * typo * Correct gini coefficient mathcal formula (#2932) * typo * typo * Update from_networkx (#2923) * Update from_networkx * Update test * Update convert.py * Minor corrections * Update test_convert.py * Corrections * Update test_convert.py * Case where there are no edges * Correcting how edge_attr are concatenated * clean up + new test * remove unused code * add union type Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * fix deterministic ordering in from_networkx * recursive-include *.jinja files Co-authored-by: Dongkwan Kim <todoaskit@gmail.com> Co-authored-by: Markus <markus.zopf@outlook.com> Co-authored-by: Jimmie <jimmiebtlr@gmail.com> Co-authored-by: Jinu Sunil <jinu.sunil@gmail.com> Co-authored-by: Moritz R Schäfer <moritz.schaefer@protonmail.com> Co-authored-by: PabloAMC <pmorenocf@alumnos.unex.es> Co-authored-by: Moritz Blum <31183934+moritzblum@users.noreply.github.com> Co-authored-by: fbragman <fbragman@users.noreply.github.com> Co-authored-by: Christopher Lee <2824685+CCInc@users.noreply.github.com> Co-authored-by: Tim Daubenschütz <tim@daubenschuetz.de> * resolve merge conflicts 3 * resolve merge conflicts 4 * Implementation of the `HGTLoader` + `ogbn-mag` example (#73) * first try * update * HGT Loader * typo * first try * update * HGT Loader * typo * bugfixes * lazy GATConv * bugfix * bugfix * full working pipeline * update * rename * docstring * typos * update * typo * typo * typo * added comments * add test * add tests * fix example * rename * linting * Random split functionalities (#72) * link split * create split * example tests * link split tests * fix linting * update docstring * undirected option, refactor and docs * add num nodes as argument to neg sampling * clean up + remove single object * update example * typo * fix compose Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * add basetransform * typo * typo * fix test * Improve `torch_geometric.data` Documentation (#98) * update data doc * typo * typo * note * typo * add docstring * only show inherited members for data and hetero_data * documentation update for batch and dataset * update doc * update * fix * record_stream * update * typo * add/fix data functionality * linting * typo * `_parent` memory leak fix (#103) * memory leak fix * Clean up * clean up * bugfix tests * typos * fix test * fix test * rename reverse * (Heterogeneous) `NeighborLoader` (#92) * initial commit * typo * neighbor loader functionality + tests * docstring * fix docstring * skip tests * fix share_memory_ * typo * typo * update example * typo * share_strategy * fix cuda calls * better print * fix size * fix print * final commit * fix * some todos * preprocessed features * fix to_undirected * more documentation * update doc * fix doc * fix doc * Add benchmark code and the example with existing graph classification examples (#93) * add benchmarking utilities * update graph classification benchmark * improve code style * add pytorch-memlab for benchmark code * skip some tests when cuda is not available * add type hint when appropriate * add seed_everything to improve code * code refactoring * code refactoring * code refactoring * code improvement * remove unnecessary dataloader import * change benchmark interface with decorator * documentation improvement * linting * linting part 2 * linting part 3 * seed_everything * create utils file * update * use utils functions * fix test * update the profiler to the latest torch (1.8.1+) * refactor profiler and add more documentation * refactor profiler and add more documentation * resolve lint errors * resolve lint errors * update * clean up test and profile * fix linting * add to doc * fix doc * typo * update benchmark Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * Move `HGTLoader` to `torch_geometric.loader` + clean up (#104) * move files * use utils functions * fix example * update * fix tests * fix seed * fix linear test * rename * Support GraphGym custom modules outside PyG package (#102) * GraphGym cleaned version * remove deepsnap dependency * fix lint errors, part 1 * fix all lint errors * fix all lint errors * fix all lint errors * apply yapf * Integrate graphgym into pyg, keep user API in project root * fix merge conflict * fix lint errors * Make optional dependencies * merge LICENSE from GraphGym * Enable adding GraphGym customized modules outside PyG package * lint * Rename `AddTrainValTestMask` to `RandomNodeSplit` (#108) * initial commit * rename example * remove AddTrainValTestMask * fix linting * create optimizer config and scheduler config separately (#113) * create optimizer config and scheduler config separately * fix format * import explicitly Co-authored-by: Dong Wang <dongwang@yannis-air.lan> * Heterogeneous Graph Tutorial (#83) * add HG tutorial roadmap * started working on hg tutorial * hg_tutorial, some text and .tex figure * added svg * hg tutorial content * fix CI * text and structure * finished first draft * fixed one code example * fixing conventions * fixing links * update svg * some smaller improvements of tutorial * improvements on tutorial * hg-tutorial: fixed compiling issue, added detailed content * added absolute links * fixed warnings * streamlined dataset section * update svg * update tutorial * update 2 Co-authored-by: Jan Eric Lenssen <janeric.lenssen@tu-dortmund.de> * typo * Move data loaders to `torch_geometric.loader` (#110) * move graphsaint * deprecations * move clusterloader * deprecations * type hints * move shadow * typo * typo * move datalistloader * dense data loader * random node sampler * fix doc * Lazy GNN operators (#89) * lazy cheb conv * lazy GraphConv * lazy GATv2Conv * lazy TAGConv * lazy FAConv * lazy FeaStConv * lazy NNConv * typo * fix tests * lazy SuperGATConv * lazy SuperGATConv fix * lazy SplineConv * fix lazy check * lazy GravNetConv * arma conv lazy * dense linear in gmmconv * typo * add test * lazy GMMConv * doc * rename (#116) * Revisit `MetaPath2Vec` (#114) * revisit metapath2vec * update * typo * update * fix doc * update * check for attributes rather than key * Clean up `torch_geometric.profile` further (#111) * remove print_layer_stats * typos * update * readme highlights and quick tour (#99) * readme highlights and quick tour * arch * arch image * arch overview * list categories * categorization * category description * Update README.md from Matthias Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * improved highlights * Update README.md Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * Update README.md Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * Update README.md Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * Update README.md Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * minor * update readme * update * update * update * update * fix url * update * update * update * update * update * update * move ops * toc * typo * typo * add svgs * update figure * fix links * fix size * fix size * typo Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * fix broken links * fix links * Heterogeneous Graph Sampler Tutorial (#117) * initial commit * address comments * remove todo * typo * Conversion between heterogenous and homogeneous graph objects (#115) * temp checkpoint (wip, will remove) * (wip) typed graph conversion * (wip) typed graph conversion * (wip) typed graph conversion * update * typo * delete examples Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de> * fix test * update doc * deprecate NeighborSampler (#119) * Move `torch_geometric.data.DataLoader` to `torch_geometric.loader.DataLoader` (#120) * move dataloader * rename * typos * typos * fix __cat_dim__ * updategp * Deprecate `train_test_split_edges` + Modifications to `RandomLinkSplit` (#121) * deprecate train_test_split_edges * to device transform * fix example * add split_labels argument * fix autoencoder example * typos * add docstring * ARGVA * seal * adress comments * Create example to load `*.csv` and transfer to `HeteroData` (#76) * create example to load csv file and transfer to heter-data * add ipython notebook version load csv with documentation * address comment * first version of csv loading doc * first version of csv loading doc * suggestion docs/source/notes/loading_csv.rst Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * suggestion docs/source/notes/loading_csv.rst Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * suggestion docs/source/notes/loading_csv.rst Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * suggestion docs/source/notes/loading_csv.rst Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * suggestions csv tutorial * example script load csv + extract fix * fixed edge index stacking dimension in example and jupyter nb * linting * linting2 * rename * update * update * update * typo * typo * update * rename * update tutorial * typo * address comments Co-authored-by: Dong Wang <dongwang@yannis-air.lan> Co-authored-by: Jan Eric Lenssen <janeric.lenssen@tu-dortmund.de> Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de> * typo * fix * typo * update * fix * fix Co-authored-by: benedekrozemberczki <benedek.rozemberczki@gmail.com> Co-authored-by: Rex Ying <rexying@stanford.edu> Co-authored-by: Dongkwan Kim <todoaskit@gmail.com> Co-authored-by: Markus <markus.zopf@outlook.com> Co-authored-by: Jimmie <jimmiebtlr@gmail.com> Co-authored-by: Jinu Sunil <jinu.sunil@gmail.com> Co-authored-by: Moritz R Schäfer <moritz.schaefer@protonmail.com> Co-authored-by: Jiaxuan <youjiaxuan@gmail.com> Co-authored-by: PabloAMC <pmorenocf@alumnos.unex.es> Co-authored-by: Moritz Blum <31183934+moritzblum@users.noreply.github.com> Co-authored-by: fbragman <fbragman@users.noreply.github.com> Co-authored-by: Christopher Lee <2824685+CCInc@users.noreply.github.com> Co-authored-by: Tim Daubenschütz <tim@daubenschuetz.de> Co-authored-by: Yue Zhao <yzhao062@gmail.com> Co-authored-by: Dong Wang <dongw89@gmail.com> Co-authored-by: Dong Wang <dongwang@yannis-air.lan> Co-authored-by: Jan Eric Lenssen <janeric.lenssen@tu-dortmund.de>
I proceeded with the installation instructions. And I have Cuda available on my machine.
However, when I run
python cora.py, I get the messageFile "/home/anaconda/envs/pytorch/lib/python3.6/site-packages/torch_scatter/utils/ext.py", line 5, in
import scatter_cuda
ModuleNotFoundError: No module named 'scatter_cuda'
I tried to clone the
scattermodule from Github separately, and then runpython setup.py installon it, with same results.I can find the
scatter_cuda.pyandscatter_cpu.pyfiles inside the "egg" namedtorch_scatter-1.0.4-py3.6-linux-x86_64.egg, and I also see the corresponding.sofiles inside that egg. But no such files inside thetorch_scatterfolder.The text was updated successfully, but these errors were encountered: