[jira] [Commented] (MXNET-36) Update ndarray binary ops to use kernel launch interface
[ https://issues.apache.org/jira/browse/MXNET-36?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389194#comment-16389194 ] ASF GitHub Bot commented on MXNET-36: - rahul003 commented on a change in pull request #9029: [MXNET-36] Update ndarray binary ops to use kernel launch instead of mshadow operations URL: https://github.com/apache/incubator-mxnet/pull/9029#discussion_r172760891 ## File path: src/ndarray/ndarray_function-inl.h ## @@ -446,6 +468,11 @@ DECL_SCALAR(DEVICE, Minus, EvalScalar_, true) DECL_SCALAR(DEVICE, Mul, EvalScalar_, true) DECL_SCALAR(DEVICE, Div, EvalScalar_, true) +DECL_BINARY_LAUNCH(DEVICE, Plus) Review comment: Removed the earlier ones, thanks This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > Update ndarray binary ops to use kernel launch interface > > > Key: MXNET-36 > URL: https://issues.apache.org/jira/browse/MXNET-36 > Project: Apache MXNet > Issue Type: New Feature >Reporter: Rahul Huilgol >Assignee: Chris Olivier >Priority: Major > > Replace older deprecated mshadow style binary operations with faster kernel > launch interface. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Created] (MXNET-35) simplify import of citation metadata
Katrin Leinweber created MXNET-35: - Summary: simplify import of citation metadata Key: MXNET-35 URL: https://issues.apache.org/jira/browse/MXNET-35 Project: Apache MXNet Issue Type: Improvement Reporter: Katrin Leinweber Assignee: Chris Olivier As suggested in [https://www.software.ac.uk/blog/2013-09-02-encouraging-citation-software-introducing-citation-files], how about adding a `CITATION` file with the BibTeX snippet of [https://arxiv.org/abs/1512.01274] to the repo? In order to make it simpler for people to cite this software properly. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-30) Add label augmentation support in image recordio2
[ https://issues.apache.org/jira/browse/MXNET-30?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389124#comment-16389124 ] ASF GitHub Bot commented on MXNET-30: - nihui commented on a change in pull request #9960: [MXNET-30] Add label augmentation support in image recordio2 URL: https://github.com/apache/incubator-mxnet/pull/9960#discussion_r172752595 ## File path: src/io/image_iter_common.h ## @@ -85,6 +85,14 @@ class ImageLabelMap { CHECK(it != idx2label_.end()) << "fail to find imagelabel for id " << imid; return mshadow::Tensor(it->second, mshadow::Shape1(label_width)); } + /*! \brief find a label for corresponding index, return vector as copy */ + inline std::vector FindCopy(size_t imid) const { Review comment: FindCopy is borrowed from the similar existing code [1], so I use FindCopy for consistency. But if you do think FindLabel is better, I could update the name, and may open another PR to update the existing FindCopy code. please tell me if I need to do this. [1] https://github.com/apache/incubator-mxnet/blob/0fe04e9d778ac9c55e933ff3f21c4ddf28a4a101/src/io/iter_image_det_recordio.cc#L116 This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > Add label augmentation support in image recordio2 > - > > Key: MXNET-30 > URL: https://issues.apache.org/jira/browse/MXNET-30 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Ni Hui >Assignee: Chris Olivier >Priority: Major > Labels: io > > pass the label data to the augmenter implementation so that the label could > be changed when image transformation applied during augmentation. > It is useful for regression task such as landmark localization. > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
[ https://issues.apache.org/jira/browse/MXNET-33?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389103#comment-16389103 ] ASF GitHub Bot commented on MXNET-33: - piiswrong commented on a change in pull request #10021: [MXNET-33] Mkldnn pooling convention crash URL: https://github.com/apache/incubator-mxnet/pull/10021#discussion_r172750399 ## File path: src/operator/nn/mkldnn/mkldnn_pooling-inl.h ## @@ -92,6 +92,8 @@ inline bool SupportMKLDNNPooling(const PoolingParam , if (param.pooling_convention == pool_enum::kValid) return true; + else +return false; Review comment: this makes the later code unreachable. Are you sure this is what you want? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > mxnet mkl-dnn crashes for Max pooling with convention "full" > - > > Key: MXNET-33 > URL: https://issues.apache.org/jira/browse/MXNET-33 > Project: Apache MXNet > Issue Type: Bug >Reporter: Ashok Emani >Assignee: Chris Olivier >Priority: Minor > > A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for > the following op. > > "op": "Pooling", > "attrs": { > "kernel": "(2, 2)", > "pool_type": "max", > "pooling_convention": "full", > "stride": "(2, 2)" > }, > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
[ https://issues.apache.org/jira/browse/MXNET-33?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389094#comment-16389094 ] ASF GitHub Bot commented on MXNET-33: - TaoLv commented on issue #10021: [MXNET-33] Mkldnn pooling convention crash URL: https://github.com/apache/incubator-mxnet/pull/10021#issuecomment-371034685 There is a test case for max pooling with `full` convention: https://github.com/apache/incubator-mxnet/blob/master/tests/python/gpu/test_operator_gpu.py#L841 But it isn't observed that this case failed before. If the test case doesn't cover the corner case you mentioned, maybe we need add new test case for it. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > mxnet mkl-dnn crashes for Max pooling with convention "full" > - > > Key: MXNET-33 > URL: https://issues.apache.org/jira/browse/MXNET-33 > Project: Apache MXNet > Issue Type: Bug >Reporter: Ashok Emani >Assignee: Chris Olivier >Priority: Minor > > A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for > the following op. > > "op": "Pooling", > "attrs": { > "kernel": "(2, 2)", > "pool_type": "max", > "pooling_convention": "full", > "stride": "(2, 2)" > }, > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
[ https://issues.apache.org/jira/browse/MXNET-33?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389065#comment-16389065 ] ASF GitHub Bot commented on MXNET-33: - ashokei commented on issue #10021: [MXNET-33] Mkldnn pooling convention crash URL: https://github.com/apache/incubator-mxnet/pull/10021#issuecomment-371031468 @zheng-da @TaoLv please review. I notice we do not have proper "full" pooling convention support, this PR disables mkl-dnn pooling for "full" convention; We will probably need a new PR to implement the "full" convention feature with mkl-dnn for both Max and Avg Pooling cases. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > mxnet mkl-dnn crashes for Max pooling with convention "full" > - > > Key: MXNET-33 > URL: https://issues.apache.org/jira/browse/MXNET-33 > Project: Apache MXNet > Issue Type: Bug >Reporter: Ashok Emani >Assignee: Chris Olivier >Priority: Minor > > A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for > the following op. > > "op": "Pooling", > "attrs": { > "kernel": "(2, 2)", > "pool_type": "max", > "pooling_convention": "full", > "stride": "(2, 2)" > }, > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
[ https://issues.apache.org/jira/browse/MXNET-33?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389048#comment-16389048 ] ASF GitHub Bot commented on MXNET-33: - ashokei opened a new pull request #10021: [MXNET-33] Mkldnn pooling convention crash URL: https://github.com/apache/incubator-mxnet/pull/10021 ## Description ## As crash is seen if pooling convention is "full" with Max pool. This PR resolves the crash by defaulting to CPU implementation when pooling convention is "full". ## Checklist ## ### Essentials ### - [ ✅ ] Passed code style checking (`make lint`) - [ ✅ ] Changes are complete (i.e. I finished coding on this PR) ### Changes ### - disable mkl-dnn if pooling convention is "full" ## Comments ## - This PR is to resolve the crash seen. A new PR may be needed to implement "full" pooling convention with mkl-dnn. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > mxnet mkl-dnn crashes for Max pooling with convention "full" > - > > Key: MXNET-33 > URL: https://issues.apache.org/jira/browse/MXNET-33 > Project: Apache MXNet > Issue Type: Bug >Reporter: Ashok Emani >Assignee: Chris Olivier >Priority: Minor > > A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for > the following op. > > "op": "Pooling", > "attrs": { > "kernel": "(2, 2)", > "pool_type": "max", > "pooling_convention": "full", > "stride": "(2, 2)" > }, > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Created] (MXNET-34) Onnx Module to import onnx models into mxnet
Rajan created MXNET-34: -- Summary: Onnx Module to import onnx models into mxnet Key: MXNET-34 URL: https://issues.apache.org/jira/browse/MXNET-34 Project: Apache MXNet Issue Type: Task Reporter: Rajan Assignee: Chris Olivier This module tries to import onnx models into mxnet. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
[ https://issues.apache.org/jira/browse/MXNET-33?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Ashok Emani updated MXNET-33: - Priority: Minor (was: Major) > mxnet mkl-dnn crashes for Max pooling with convention "full" > - > > Key: MXNET-33 > URL: https://issues.apache.org/jira/browse/MXNET-33 > Project: Apache MXNet > Issue Type: Bug >Reporter: Ashok Emani >Assignee: Chris Olivier >Priority: Minor > > A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for > the following op. > > "op": "Pooling", > "attrs": { > "kernel": "(2, 2)", > "pool_type": "max", > "pooling_convention": "full", > "stride": "(2, 2)" > }, > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Created] (MXNET-33) mxnet mkl-dnn crashes for Max pooling with convention "full"
Ashok Emani created MXNET-33: Summary: mxnet mkl-dnn crashes for Max pooling with convention "full" Key: MXNET-33 URL: https://issues.apache.org/jira/browse/MXNET-33 Project: Apache MXNet Issue Type: Bug Reporter: Ashok Emani Assignee: Chris Olivier A crash is seen when running SSD-VGG16 300 VOC , with mxnet mkldnn build, for the following op. "op": "Pooling", "attrs": { "kernel": "(2, 2)", "pool_type": "max", "pooling_convention": "full", "stride": "(2, 2)" }, -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Created] (MXNET-32) Improve documentation of environment variables
Rahul Huilgol created MXNET-32: -- Summary: Improve documentation of environment variables Key: MXNET-32 URL: https://issues.apache.org/jira/browse/MXNET-32 Project: Apache MXNet Issue Type: Improvement Reporter: Rahul Huilgol Assignee: Chris Olivier -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Commented] (MXNET-30) Add label augmentation support in image recordio2
[ https://issues.apache.org/jira/browse/MXNET-30?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16389034#comment-16389034 ] ASF GitHub Bot commented on MXNET-30: - nihui commented on issue #9960: [MXNET-30] Add label augmentation support in image recordio2 URL: https://github.com/apache/incubator-mxnet/pull/9960#issuecomment-371023666 jenkins please retest changes This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > Add label augmentation support in image recordio2 > - > > Key: MXNET-30 > URL: https://issues.apache.org/jira/browse/MXNET-30 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Ni Hui >Assignee: Chris Olivier >Priority: Major > Labels: io > > pass the label data to the augmenter implementation so that the label could > be changed when image transformation applied during augmentation. > It is useful for regression task such as landmark localization. > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Created] (MXNET-30) Add label augmentation support in image recordio2
Ni Hui created MXNET-30: --- Summary: Add label augmentation support in image recordio2 Key: MXNET-30 URL: https://issues.apache.org/jira/browse/MXNET-30 Project: Apache MXNet Issue Type: Improvement Reporter: Ni Hui Assignee: Chris Olivier pass the label data to the augmenter implementation so that the label could be changed when image transformation applied during augmentation. It is useful for regression task such as landmark localization. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-29) FIx profiler multiple start/stop in same process
[ https://issues.apache.org/jira/browse/MXNET-29?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-29: --- Component/s: MXNet Engine > FIx profiler multiple start/stop in same process > > > Key: MXNET-29 > URL: https://issues.apache.org/jira/browse/MXNET-29 > Project: Apache MXNet > Issue Type: Bug > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: profiler > > Right now, after the first file is written, > On stop, it needs to flush and wait > On start, probably needs to reset num records and maybe other vars > Something else? -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-28) SPIKE: tf.assign equivalent or alternative
[ https://issues.apache.org/jira/browse/MXNET-28?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-28: --- Component/s: MXNet Engine Keras > SPIKE: tf.assign equivalent or alternative > -- > > Key: MXNET-28 > URL: https://issues.apache.org/jira/browse/MXNET-28 > Project: Apache MXNet > Issue Type: Task > Components: Keras, MXNet Engine >Reporter: Chris Olivier >Priority: Major > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-29) FIx profiler multiple start/stop in same process
[ https://issues.apache.org/jira/browse/MXNET-29?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-29: --- Labels: profiler (was: ) > FIx profiler multiple start/stop in same process > > > Key: MXNET-29 > URL: https://issues.apache.org/jira/browse/MXNET-29 > Project: Apache MXNet > Issue Type: Bug > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: profiler > > Right now, after the first file is written, > On stop, it needs to flush and wait > On start, probably needs to reset num records and maybe other vars > Something else? -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-26) MXNet Keras 2.0 Support
[ https://issues.apache.org/jira/browse/MXNET-26?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-26: --- Component/s: Keras > MXNet Keras 2.0 Support > --- > > Key: MXNET-26 > URL: https://issues.apache.org/jira/browse/MXNET-26 > Project: Apache MXNet > Issue Type: Epic > Components: Keras >Reporter: Chris Olivier >Priority: Major > Labels: SDK, keras > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-25) Convert created profiling objects to use ObjectPool
[ https://issues.apache.org/jira/browse/MXNET-25?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-25: --- Component/s: MXNet Engine > Convert created profiling objects to use ObjectPool > --- > > Key: MXNET-25 > URL: https://issues.apache.org/jira/browse/MXNET-25 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance, profliling > > In order to avoid using new/delete in performance-sensitive execution paths, > some profiling objects, such as ProfileOperator might be better to use > ObjectPool as their memory/object manager > Ideally, this could occur by overloading the new/delete operators in a base > class (templated or nor), which would then call ObjectPool.New() and > ObjectPool.Delete() -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-25) Convert created profiling objects to use ObjectPool
[ https://issues.apache.org/jira/browse/MXNET-25?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-25: --- Labels: performance profliling (was: ) > Convert created profiling objects to use ObjectPool > --- > > Key: MXNET-25 > URL: https://issues.apache.org/jira/browse/MXNET-25 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance, profliling > > In order to avoid using new/delete in performance-sensitive execution paths, > some profiling objects, such as ProfileOperator might be better to use > ObjectPool as their memory/object manager > Ideally, this could occur by overloading the new/delete operators in a base > class (templated or nor), which would then call ObjectPool.New() and > ObjectPool.Delete() -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-24) Cython build and deployment for Makefile and/or pip install, setup.py or some other means
[ https://issues.apache.org/jira/browse/MXNET-24?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-24: --- Component/s: MXNet Engine CI Build > Cython build and deployment for Makefile and/or pip install, setup.py or some > other means > - > > Key: MXNET-24 > URL: https://issues.apache.org/jira/browse/MXNET-24 > Project: Apache MXNet > Issue Type: Task > Components: CI Build , MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: cython > > Close the loop on Cython build and deployment. > May mean that Makefile calls cmake, or setup.py uses cython tools, or some > other mechanism -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-21) Refactor kvstore-invoked python optimizer code to C++ and cython
[ https://issues.apache.org/jira/browse/MXNET-21?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-21: --- Labels: performance (was: ) > Refactor kvstore-invoked python optimizer code to C++ and cython > > > Key: MXNET-21 > URL: https://issues.apache.org/jira/browse/MXNET-21 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance > > KVStore calls back into python for the optimizer pass. This is known to be a > performance bottleneck for multi-GPU and distributed training -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-19) Gluon Performance
[ https://issues.apache.org/jira/browse/MXNET-19?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-19: --- Labels: gluon performance (was: ) > Gluon Performance > - > > Key: MXNET-19 > URL: https://issues.apache.org/jira/browse/MXNET-19 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: gluon, performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-20) Cache Variables
[ https://issues.apache.org/jira/browse/MXNET-20?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-20: --- Labels: gluon perfomance (was: ) > Cache Variables > --- > > Key: MXNET-20 > URL: https://issues.apache.org/jira/browse/MXNET-20 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: gluon, perfomance > > Add ability to cache graph variabled between graph executions in hybrid mode. > Currently, it is believed that this causes the major performance difference > between symbolic and gluon modes -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-19) Gluon Performance
[ https://issues.apache.org/jira/browse/MXNET-19?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-19: --- Component/s: MXNet Engine > Gluon Performance > - > > Key: MXNET-19 > URL: https://issues.apache.org/jira/browse/MXNET-19 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: gluon, performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-15) Resolve many-variable create and destroy bottleneck in gluon dynamic graph execution
[ https://issues.apache.org/jira/browse/MXNET-15?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-15: --- Labels: gluon performance (was: ) > Resolve many-variable create and destroy bottleneck in gluon dynamic graph > execution > > > Key: MXNET-15 > URL: https://issues.apache.org/jira/browse/MXNET-15 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: gluon, performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-16) Move submodules to the 3rdparty folder
[ https://issues.apache.org/jira/browse/MXNET-16?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-16: --- Component/s: CI Build > Move submodules to the 3rdparty folder > -- > > Key: MXNET-16 > URL: https://issues.apache.org/jira/browse/MXNET-16 > Project: Apache MXNet > Issue Type: Task > Components: CI Build >Reporter: Haibin Lin >Priority: Major > Labels: 3rdParty > > MXNet depends on many submodules from [dmlc|https://github.com/dmlc/] > including nnvm, ps-lite, mshadow, dlpack and dmlc-core. These submodules are > not yet moved into the 3rdparty folder, which creates confusion when > reviewing the licenses. > We need to move all these submodules to the 3rdparty folder. Specifically, we > need to change both cmake and make build scripts, for multiple build target > (MXNet core, cpp-package, amalgamation, etc) and make sure it is tested in > multiple environments. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-16) Move submodules to the 3rdparty folder
[ https://issues.apache.org/jira/browse/MXNET-16?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-16: --- Labels: 3rdParty (was: ) > Move submodules to the 3rdparty folder > -- > > Key: MXNET-16 > URL: https://issues.apache.org/jira/browse/MXNET-16 > Project: Apache MXNet > Issue Type: Task > Components: CI Build >Reporter: Haibin Lin >Priority: Major > Labels: 3rdParty > > MXNet depends on many submodules from [dmlc|https://github.com/dmlc/] > including nnvm, ps-lite, mshadow, dlpack and dmlc-core. These submodules are > not yet moved into the 3rdparty folder, which creates confusion when > reviewing the licenses. > We need to move all these submodules to the 3rdparty folder. Specifically, we > need to change both cmake and make build scripts, for multiple build target > (MXNet core, cpp-package, amalgamation, etc) and make sure it is tested in > multiple environments. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-15) Resolve many-variable create and destroy bottleneck in gluon dynamic graph execution
[ https://issues.apache.org/jira/browse/MXNET-15?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-15: --- Component/s: MXNet Engine > Resolve many-variable create and destroy bottleneck in gluon dynamic graph > execution > > > Key: MXNET-15 > URL: https://issues.apache.org/jira/browse/MXNET-15 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: gluon, performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-11) Multithreaded Inference
[ https://issues.apache.org/jira/browse/MXNET-11?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-11: --- Component/s: MXNet Engine > Multithreaded Inference > --- > > Key: MXNET-11 > URL: https://issues.apache.org/jira/browse/MXNET-11 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: inference > > Add the ability to do multithreaded inference without using fork() or using > multiple copies of a given model -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-14) Profiling gluon Multi-GPU performance bottlenecks
[ https://issues.apache.org/jira/browse/MXNET-14?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-14: --- Component/s: MXNet Engine > Profiling gluon Multi-GPU performance bottlenecks > - > > Key: MXNET-14 > URL: https://issues.apache.org/jira/browse/MXNET-14 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-11) Multithreaded Inference
[ https://issues.apache.org/jira/browse/MXNET-11?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-11: --- Labels: inference (was: ) > Multithreaded Inference > --- > > Key: MXNET-11 > URL: https://issues.apache.org/jira/browse/MXNET-11 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: inference > > Add the ability to do multithreaded inference without using fork() or using > multiple copies of a given model -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-12) Spike: Investigate current limitations
[ https://issues.apache.org/jira/browse/MXNET-12?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-12: --- Component/s: MXNet Engine > Spike: Investigate current limitations > -- > > Key: MXNET-12 > URL: https://issues.apache.org/jira/browse/MXNET-12 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: inference, performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-14) Profiling gluon Multi-GPU performance bottlenecks
[ https://issues.apache.org/jira/browse/MXNET-14?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-14: --- Labels: performance (was: ) > Profiling gluon Multi-GPU performance bottlenecks > - > > Key: MXNET-14 > URL: https://issues.apache.org/jira/browse/MXNET-14 > Project: Apache MXNet > Issue Type: Task > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance > -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-13) Add attributes to profiling summary data/operator names
[ https://issues.apache.org/jira/browse/MXNET-13?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-13: --- Component/s: MXNet Engine > Add attributes to profiling summary data/operator names > --- > > Key: MXNET-13 > URL: https://issues.apache.org/jira/browse/MXNET-13 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance, profliling > > In profiling object naming, include some attributes, such as input and > outputs shapes > > Some discussion can be found here: > https://github.com/apache/incubator-mxnet/issues/9545 -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-8) Add basic cython build to cmake build system
[ https://issues.apache.org/jira/browse/MXNET-8?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-8: -- Component/s: MXNet Engine > Add basic cython build to cmake build system > > > Key: MXNET-8 > URL: https://issues.apache.org/jira/browse/MXNET-8 > Project: Apache MXNet > Issue Type: Improvement > Components: CI Build , MXNet Engine >Reporter: Chris Olivier >Assignee: Chris Olivier >Priority: Major > Labels: cython, performance > > Make Cython code buildable and runnable with CMake builds. > Will address Makefile or other build types in subsequent ticket(s) -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-10) General Performance
[ https://issues.apache.org/jira/browse/MXNET-10?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-10: --- Labels: performance (was: ) > General Performance > --- > > Key: MXNET-10 > URL: https://issues.apache.org/jira/browse/MXNET-10 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance > > General Performance Epic -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-10) General Performance
[ https://issues.apache.org/jira/browse/MXNET-10?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-10: --- Component/s: MXNet Engine > General Performance > --- > > Key: MXNET-10 > URL: https://issues.apache.org/jira/browse/MXNET-10 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: performance > > General Performance Epic -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-4) Refactor Random and ParallelRandom resources to use MKL for MKL builds
[ https://issues.apache.org/jira/browse/MXNET-4?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-4: -- Component/s: MXNet Engine > Refactor Random and ParallelRandom resources to use MKL for MKL builds > -- > > Key: MXNET-4 > URL: https://issues.apache.org/jira/browse/MXNET-4 > Project: Apache MXNet > Issue Type: Improvement > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: mkl, performance > > Refactor Random and ParallelRandom resources to use MKL for MKL builds > Things such as RngUniform, etc. Similarly to what is done for dropout > operator. > It may need to allocate some temporary memory and generate random numbers in > batches, then serving them out from that batch. > Also the Random classes could export a "fill buffer with randoms" function, > which seems to be a common use-case and fits the MKL API more closely. > Care must be taken regarding MKL's fixed output types for some of the API > functions. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
[jira] [Updated] (MXNET-6) Cython Refactoring
[ https://issues.apache.org/jira/browse/MXNET-6?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Olivier updated MXNET-6: -- Component/s: MXNet Engine > Cython Refactoring > -- > > Key: MXNET-6 > URL: https://issues.apache.org/jira/browse/MXNET-6 > Project: Apache MXNet > Issue Type: Epic > Components: MXNet Engine >Reporter: Chris Olivier >Priority: Major > Labels: cython, performance > > Refactor some of the python code to cython for performance reasons -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org