[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345617102 For the text part, you're using end-of-sentence token as the padding, and are unrolling all the way even for paddings. Usually you would want to 1) use a separate token for padding, and 2) minimize the impact of padding by including as little of it as possible. Having these long paddings can impact your text model performance. 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345616379 From the screenshot it looks like the indexing of the text is correct. Do the questions match the input images? 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345615703 I'm asking you to try and overfit a small dataset first as sanity check. If your model can't overfit a small subset of the data, then your model is wrong. This is the first step. Unfortunately I can't read the paper and reproduce this for you right now. 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 With regards, Apache Git Services
[GitHub] liuzhi136 commented on issue #8720: Implementation Help!!!
liuzhi136 commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345613723 I still wander whether the way I pad end character is correct ? 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 With regards, Apache Git Services
[GitHub] liuzhi136 commented on issue #8720: Implementation Help!!!
liuzhi136 commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345611895 I just examine the data batch I provide to the model. one of batches looks like below and the next one is its corresponding text version. I can assure you that my data pipeline is working correctly. The data set I use to train the model is the same as used in this paper, and I just want to reproduce this experiment result see whether it really work. I think it is not overfitting cause the training loss and validations loss are both high. I think it's still underfitting. Can you just read this paper to check if my model structure is correct? Cause it is not too complicated and won't take you too much time. maybe about ten minutes at most. ![image](https://user-images.githubusercontent.com/13534043/33006930-47302874-ce07-11e7-95a0-f63d539f30b9.png) ![image](https://user-images.githubusercontent.com/13534043/33006123-19511100-ce04-11e7-9867-7e856d59255c.png) 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 With regards, Apache Git Services
[GitHub] liuzhi136 commented on issue #8720: Implementation Help!!!
liuzhi136 commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345611895 I just examine the data batch I provide to the model. one of batches looks like below and the next one is its corresponding text version. I can assure you that my data pipeline is working correctly. The data set I use to train the model is the same as used in this paper, and I just want to reproduce this experiment result see whether it really work. I think it is not overfitting cause the training loss and validations loss are both high. I think it's still underfitting. Can you just read this paper to check if my model structure is correct? Cause it is not too complicated and won't take you too much time. maybe about ten minutes at most. ![image](https://user-images.githubusercontent.com/13534043/33006782-a80a5e0e-ce06-11e7-8af3-e6cf6934d987.png) ![image](https://user-images.githubusercontent.com/13534043/33006123-19511100-ce04-11e7-9867-7e856d59255c.png) 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 With regards, Apache Git Services
[GitHub] MDR-EX1000 closed issue #8357: USE_CUDNN=1
MDR-EX1000 closed issue #8357: USE_CUDNN=1 URL: https://github.com/apache/incubator-mxnet/issues/8357 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 With regards, Apache Git Services
[GitHub] MDR-EX1000 closed issue #8188: Periodic Loss Value when training with "step" learning rate policy
MDR-EX1000 closed issue #8188: Periodic Loss Value when training with "step" learning rate policy URL: https://github.com/apache/incubator-mxnet/issues/8188 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8689: Mark tests that should only be run nightly.
szha commented on issue #8689: Mark tests that should only be run nightly. URL: https://github.com/apache/incubator-mxnet/pull/8689#issuecomment-345609106 @KellenSunderland please rebase and resolve conflict. 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 With regards, Apache Git Services
[GitHub] anirudh2290 commented on issue #8721: fix custom op for backward compatibility
anirudh2290 commented on issue #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721#issuecomment-345591574 Thank you for fixing this and adding the tests! Yes, since the output stypes which is the in_grads would depend on arguments length and not output length. Can you please let me know what LR example you are referring to? 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 With regards, Apache Git Services
[GitHub] ZiyueHuang commented on issue #8721: fix custom op for backward compatibility
ZiyueHuang commented on issue #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721#issuecomment-345605234 @anirudh2290 Thanks for your comments. LR example is the `linear_classification.py` in example/sparse. 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345602238 I'm not familiar with this model, so I'm trying to help by offering the perspective of how I would approach it. Have you tried anything to find which parts of the whole pipeline might have problems? For example, are you sure your data pipeline is working correctly? Have you examined samples from your iterator to see if it looks correct? If just using a small subset as training set, does your model overfit to the set? 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345602238 I'm not familiar with this model, so I'm trying to help by offering from the perspective of how I would approach it. Have you tried anything to find which parts of the whole pipeline might have problems? For example, are you sure your data pipeline is working correctly? Have you examined samples from your iterator to see if it looks correct? If just using a small subset as training set, does your model overfit to the set? 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 With regards, Apache Git Services
[GitHub] dwSun commented on issue #8693: mx.image.ImageIter failed on lst files failed with " src/io/image_io.cc:168: Check failed: static_cast<void*>(dst.ptr()) == out->data().dptr"
dwSun commented on issue #8693: mx.image.ImageIter failed on lst files failed with " src/io/image_io.cc:168: Check failed: static_cast(dst.ptr()) == out->data().dptr" URL: https://github.com/apache/incubator-mxnet/issues/8693#issuecomment-345602144 I am a beginner on mxnet. No idea what I should do when face with such kind of problem. According to document of **mxnet.image.ImageIter**: >To load input images from .rec files, use path_imgrec parameter and to load from raw image files, use path_imglist and path_root parameters. I can load raw image files directly without rec files, that will save me some time when debug my model. > the imdecoding is always performed on cpu at this time. Is this documented some where? I can't find 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 With regards, Apache Git Services
[GitHub] KeyKy commented on issue #8708: dataloader._batchify function is slow
KeyKy commented on issue #8708: dataloader._batchify function is slow URL: https://github.com/apache/incubator-mxnet/issues/8708#issuecomment-345599857 37 118 5568 47.2 3.2 data = np.asarray(data) 38 11839818337.4 23.0 return nd.array(data, dtype=data.dtype in my project, I also find np to nd is slow. To my knowledge, python numpy array to C++ is zero cost. 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 With regards, Apache Git Services
[GitHub] KeyKy closed issue #8538: Can i fix bias to 0 and gbias also 0 in BatchNorm?
KeyKy closed issue #8538: Can i fix bias to 0 and gbias also 0 in BatchNorm? URL: https://github.com/apache/incubator-mxnet/issues/8538 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 With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #8704: Initial Prep for 1.0: bump up version and add 0.12.1 changes to master
szha commented on a change in pull request #8704: Initial Prep for 1.0: bump up version and add 0.12.1 changes to master URL: https://github.com/apache/incubator-mxnet/pull/8704#discussion_r151907672 ## File path: docs/build_version_doc/build_all_version.sh ## @@ -21,7 +21,7 @@ # Built files are stored in $built # Version numbers are stored in $tag_list. # Version numbers are ordered from latest to old and final one is master. -tag_list="0.12.0 0.11.0 master" +tag_list="1.0.0 0.12.0 0.11.0 master" Review comment: 1.0.0 is not available yet. 0.12.1 should be added instead. 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 With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #8704: Initial Prep for 1.0: bump up version and add 0.12.1 changes to master
szha commented on a change in pull request #8704: Initial Prep for 1.0: bump up version and add 0.12.1 changes to master URL: https://github.com/apache/incubator-mxnet/pull/8704#discussion_r151907710 ## File path: setup-utils/install-mxnet-osx-python.sh ## @@ -33,7 +33,7 @@ then # TODO: Change this to latest tag # to avoid updating this value for every release # - export MXNET_TAG="0.12.0" + export MXNET_TAG="1.0.0" Review comment: 0.12.1 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 With regards, Apache Git Services
[GitHub] piiswrong closed pull request #8718: Doc src and fix
piiswrong closed pull request #8718: Doc src and fix URL: https://github.com/apache/incubator-mxnet/pull/8718 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/docs/api/python/gluon/model_zoo.md b/docs/api/python/gluon/model_zoo.md index 18d9ae2cb8..727aeb7fd8 100644 --- a/docs/api/python/gluon/model_zoo.md +++ b/docs/api/python/gluon/model_zoo.md @@ -13,6 +13,7 @@ This document lists the model APIs in Gluon: :nosignatures: mxnet.gluon.model_zoo +mxnet.gluon.model_zoo.vision ``` The `Gluon Model Zoo` API, defined in the `gluon.model_zoo` package, provides pre-defined @@ -186,6 +187,8 @@ In the rest of this document, we list routines provided by the `gluon.model_zoo` ```eval_rst +.. automodule:: mxnet.gluon.model_zoo + .. automodule:: mxnet.gluon.model_zoo.vision :members: :imported-members: diff --git a/docs/api/python/ndarray/ndarray.md b/docs/api/python/ndarray/ndarray.md index 68f8333ada..59ca4a612e 100644 --- a/docs/api/python/ndarray/ndarray.md +++ b/docs/api/python/ndarray/ndarray.md @@ -559,13 +559,13 @@ The `ndarray` package provides several classes: .. autosummary:: :nosignatures: -mxnet.nd.random.uniform -mxnet.nd.random.normal -mxnet.nd.random.gamma -mxnet.nd.random.exponential -mxnet.nd.random.poisson -mxnet.nd.random.negative_binomial -mxnet.nd.random.generalized_negative_binomial +mxnet.ndarray.random.uniform +mxnet.ndarray.random.normal +mxnet.ndarray.random.gamma +mxnet.ndarray.random.exponential +mxnet.ndarray.random.poisson +mxnet.ndarray.random.negative_binomial +mxnet.ndarray.random.generalized_negative_binomial mxnet.random.seed ``` diff --git a/docs/api/python/symbol/symbol.md b/docs/api/python/symbol/symbol.md index 8d83086fde..e383597236 100644 --- a/docs/api/python/symbol/symbol.md +++ b/docs/api/python/symbol/symbol.md @@ -558,13 +558,13 @@ Composite multiple symbols into a new one by an operator. .. autosummary:: :nosignatures: -mxnet.sym.random.uniform -mxnet.sym.random.normal -mxnet.sym.random.gamma -mxnet.sym.random.exponential -mxnet.sym.random.poisson -mxnet.sym.random.negative_binomial -mxnet.sym.random.generalized_negative_binomial +mxnet.symbol.random.uniform +mxnet.symbol.random.normal +mxnet.symbol.random.gamma +mxnet.symbol.random.exponential +mxnet.symbol.random.poisson +mxnet.symbol.random.negative_binomial +mxnet.symbol.random.generalized_negative_binomial mxnet.random.seed ``` diff --git a/docs/conf.py b/docs/conf.py index ad51323f01..d018408d45 100644 --- a/docs/conf.py +++ b/docs/conf.py @@ -59,6 +59,7 @@ 'sphinx.ext.autosummary', 'sphinx.ext.napoleon', 'sphinx.ext.mathjax', +'sphinx.ext.viewcode', 'breathe', 'mxdoc' ] diff --git a/docs/mxdoc.py b/docs/mxdoc.py index 26e4c9e265..caf135680d 100644 --- a/docs/mxdoc.py +++ b/docs/mxdoc.py @@ -62,8 +62,12 @@ def generate_doxygen(app): def build_mxnet(app): """Build mxnet .so lib""" -_run_cmd("cd %s/.. && cp make/config.mk config.mk && make -j$(nproc) DEBUG=1" % -app.builder.srcdir) +if not os.path.exists(os.path.join(app.builder.srcdir, '..', 'config.mk')): +_run_cmd("cd %s/.. && cp make/config.mk config.mk && make -j$(nproc) DEBUG=1" % +app.builder.srcdir) +else: +_run_cmd("cd %s/.. && make -j$(nproc) DEBUG=1" % +app.builder.srcdir) def build_r_docs(app): """build r pdf""" diff --git a/python/mxnet/gluon/model_zoo/vision/__init__.py b/python/mxnet/gluon/model_zoo/vision/__init__.py index a9a539bf20..619711e71d 100644 --- a/python/mxnet/gluon/model_zoo/vision/__init__.py +++ b/python/mxnet/gluon/model_zoo/vision/__init__.py @@ -30,6 +30,7 @@ - `MobileNet`_ You can construct a model with random weights by calling its constructor: + .. code:: from mxnet.gluon.model_zoo import vision @@ -39,8 +40,8 @@ densenet = vision.densenet_161() We provide pre-trained models for all the models except ResNet V2. -These can constructed by passing -``pretrained=True``: +These can constructed by passing ``pretrained=True``: + .. code:: from mxnet.gluon.model_zoo import vision 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: Doc src and fix (#8718)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new fd45517 Doc src and fix (#8718) fd45517 is described below commit fd45517614842bfa1d32d1ba54a200eb4a0dd377 Author: Sheng ZhaAuthorDate: Sun Nov 19 21:57:15 2017 -0800 Doc src and fix (#8718) * add viewcode * fix formatting --- docs/api/python/gluon/model_zoo.md | 3 +++ docs/api/python/ndarray/ndarray.md | 14 +++--- docs/api/python/symbol/symbol.md| 14 +++--- docs/conf.py| 1 + docs/mxdoc.py | 8 ++-- python/mxnet/gluon/model_zoo/vision/__init__.py | 5 +++-- 6 files changed, 27 insertions(+), 18 deletions(-) diff --git a/docs/api/python/gluon/model_zoo.md b/docs/api/python/gluon/model_zoo.md index a22b437..8310461 100644 --- a/docs/api/python/gluon/model_zoo.md +++ b/docs/api/python/gluon/model_zoo.md @@ -13,6 +13,7 @@ This document lists the model APIs in Gluon: :nosignatures: mxnet.gluon.model_zoo +mxnet.gluon.model_zoo.vision ``` The `Gluon Model Zoo` API, defined in the `gluon.model_zoo` package, provides pre-defined @@ -182,6 +183,8 @@ In the rest of this document, we list routines provided by the `gluon.model_zoo` ```eval_rst +.. automodule:: mxnet.gluon.model_zoo + .. automodule:: mxnet.gluon.model_zoo.vision :members: :imported-members: diff --git a/docs/api/python/ndarray/ndarray.md b/docs/api/python/ndarray/ndarray.md index 68f8333..59ca4a6 100644 --- a/docs/api/python/ndarray/ndarray.md +++ b/docs/api/python/ndarray/ndarray.md @@ -559,13 +559,13 @@ The `ndarray` package provides several classes: .. autosummary:: :nosignatures: -mxnet.nd.random.uniform -mxnet.nd.random.normal -mxnet.nd.random.gamma -mxnet.nd.random.exponential -mxnet.nd.random.poisson -mxnet.nd.random.negative_binomial -mxnet.nd.random.generalized_negative_binomial +mxnet.ndarray.random.uniform +mxnet.ndarray.random.normal +mxnet.ndarray.random.gamma +mxnet.ndarray.random.exponential +mxnet.ndarray.random.poisson +mxnet.ndarray.random.negative_binomial +mxnet.ndarray.random.generalized_negative_binomial mxnet.random.seed ``` diff --git a/docs/api/python/symbol/symbol.md b/docs/api/python/symbol/symbol.md index 8d83086..e383597 100644 --- a/docs/api/python/symbol/symbol.md +++ b/docs/api/python/symbol/symbol.md @@ -558,13 +558,13 @@ Composite multiple symbols into a new one by an operator. .. autosummary:: :nosignatures: -mxnet.sym.random.uniform -mxnet.sym.random.normal -mxnet.sym.random.gamma -mxnet.sym.random.exponential -mxnet.sym.random.poisson -mxnet.sym.random.negative_binomial -mxnet.sym.random.generalized_negative_binomial +mxnet.symbol.random.uniform +mxnet.symbol.random.normal +mxnet.symbol.random.gamma +mxnet.symbol.random.exponential +mxnet.symbol.random.poisson +mxnet.symbol.random.negative_binomial +mxnet.symbol.random.generalized_negative_binomial mxnet.random.seed ``` diff --git a/docs/conf.py b/docs/conf.py index ad51323..d018408 100644 --- a/docs/conf.py +++ b/docs/conf.py @@ -59,6 +59,7 @@ extensions = [ 'sphinx.ext.autosummary', 'sphinx.ext.napoleon', 'sphinx.ext.mathjax', +'sphinx.ext.viewcode', 'breathe', 'mxdoc' ] diff --git a/docs/mxdoc.py b/docs/mxdoc.py index 26e4c9e..caf1356 100644 --- a/docs/mxdoc.py +++ b/docs/mxdoc.py @@ -62,8 +62,12 @@ def generate_doxygen(app): def build_mxnet(app): """Build mxnet .so lib""" -_run_cmd("cd %s/.. && cp make/config.mk config.mk && make -j$(nproc) DEBUG=1" % -app.builder.srcdir) +if not os.path.exists(os.path.join(app.builder.srcdir, '..', 'config.mk')): +_run_cmd("cd %s/.. && cp make/config.mk config.mk && make -j$(nproc) DEBUG=1" % +app.builder.srcdir) +else: +_run_cmd("cd %s/.. && make -j$(nproc) DEBUG=1" % +app.builder.srcdir) def build_r_docs(app): """build r pdf""" diff --git a/python/mxnet/gluon/model_zoo/vision/__init__.py b/python/mxnet/gluon/model_zoo/vision/__init__.py index a9a539b..619711e 100644 --- a/python/mxnet/gluon/model_zoo/vision/__init__.py +++ b/python/mxnet/gluon/model_zoo/vision/__init__.py @@ -30,6 +30,7 @@ This module contains definitions for the following model architectures: - `MobileNet`_ You can construct a model with random weights by calling its constructor: + .. code:: from mxnet.gluon.model_zoo import vision @@ -39,8 +40,8 @@ You can construct a model with random weights by calling its constructor: densenet = vision.densenet_161() We provide pre-trained models for
[GitHub] piiswrong closed pull request #8710: Remove experimental warning on Gluon and add Gluon tutorials
piiswrong closed pull request #8710: Remove experimental warning on Gluon and add Gluon tutorials URL: https://github.com/apache/incubator-mxnet/pull/8710 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/docs/api/python/autograd/autograd.md b/docs/api/python/autograd/autograd.md index de8188446b..410d6a94e2 100644 --- a/docs/api/python/autograd/autograd.md +++ b/docs/api/python/autograd/autograd.md @@ -1,14 +1,9 @@ # Autograd Package - ```eval_rst .. currentmodule:: mxnet.autograd ``` -```eval_rst -.. warning:: This package is currently experimental and may change in the near future. -``` - ## Overview The `autograd` package enables automatic diff --git a/docs/api/python/gluon/data.md b/docs/api/python/gluon/data.md index f72f3cd03f..0b5f959e32 100644 --- a/docs/api/python/gluon/data.md +++ b/docs/api/python/gluon/data.md @@ -15,10 +15,6 @@ This document lists the data APIs in Gluon: The `Gluon Data` API, defined in the `gluon.data` package, provides useful dataset loading and processing tools, as well as common public datasets. -```eval_rst -.. warning:: This package contains experimental APIs and may change in the near future. -``` - In the rest of this document, we list routines provided by the `gluon.data` package. ## Data diff --git a/docs/api/python/gluon/gluon.md b/docs/api/python/gluon/gluon.md index 0ef6dbed0e..2ae766fdcb 100644 --- a/docs/api/python/gluon/gluon.md +++ b/docs/api/python/gluon/gluon.md @@ -5,10 +5,6 @@ .. currentmodule:: mxnet.gluon ``` -```eval_rst -.. warning:: This package is currently experimental and may change in the near future. -``` - ## Overview diff --git a/docs/api/python/gluon/model_zoo.md b/docs/api/python/gluon/model_zoo.md index 18d9ae2cb8..a22b437640 100644 --- a/docs/api/python/gluon/model_zoo.md +++ b/docs/api/python/gluon/model_zoo.md @@ -18,10 +18,6 @@ This document lists the model APIs in Gluon: The `Gluon Model Zoo` API, defined in the `gluon.model_zoo` package, provides pre-defined and pre-trained models to help bootstrap machine learning applications. -```eval_rst -.. warning:: This package contains experimental APIs and may change in the near future. -``` - In the rest of this document, we list routines provided by the `gluon.model_zoo` package. ### Vision diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 40c381..6429dfb31b 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -2,13 +2,37 @@ These tutorials introduce a few fundamental concepts in deep learning and how to implement them in _MXNet_. The _Basics_ section contains tutorials on manipulating arrays, building networks, loading/preprocessing data, etc. The _Training and Inference_ section talks about implementing Linear Regression, training a Handwritten digit classifier using MLP and CNN, running inferences using a pre-trained model, and lastly, efficiently training a large scale image classifier. -```eval_rst -.. Note:: We are working on a set of tutorials for the new imperative interface called Gluon. A preview version is hosted at http://gluon.mxnet.io. -``` -## Python +## Gluon + +Gluon is the high-level interface for MXNet. It is more intuitive and easier to use than the lower level interface. +Gluon supports dynamic (define-by-run) graphs with JIT-compilation to achieve both flexibility and efficiency. +This is a selected subset of Gluon tutorials. For the comprehensive tutorial on Gluon, +please see [gluon.mxnet.io](http://gluon.mxnet.io). + +### Basics + +- [Manipulate data the MXNet way with ndarray](http://gluon.mxnet.io/chapter01_crashcourse/ndarray.html) +- [Automatic differentiation with autograd](http://gluon.mxnet.io/chapter01_crashcourse/autograd.html) +- [Linear regression with gluon](http://gluon.mxnet.io/chapter02_supervised-learning/linear-regression-gluon.html) +- [Serialization - saving, loading and checkpointing](http://gluon.mxnet.io/chapter03_deep-neural-networks/serialization.html) + +### Neural Networks + +- [Multilayer perceptrons in gluon](http://gluon.mxnet.io/chapter03_deep-neural-networks/mlp-gluon.html) +- [Convolutional Neural Networks in gluon](http://gluon.mxnet.io/chapter04_convolutional-neural-networks/cnn-gluon.html) +- [Recurrent Neural Networks with gluon](http://gluon.mxnet.io/chapter05_recurrent-neural-networks/rnns-gluon.html) + +### Advanced + +- [Plumbing: A look under the hood of gluon](http://gluon.mxnet.io/chapter03_deep-neural-networks/plumbing.html) +- [Designing a custom layer with gluon](http://gluon.mxnet.io/chapter03_deep-neural-networks/custom-layer.html) +- [Fast, portable neural networks with Gluon HybridBlocks](http://gluon.mxnet.io/chapter07_distributed-learning/hybridize.html) +-
[incubator-mxnet] branch master updated: Remove experimental warning on Gluon and add Gluon tutorials (#8710)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new defd3c3 Remove experimental warning on Gluon and add Gluon tutorials (#8710) defd3c3 is described below commit defd3c396831aace17bdfea0fd8913a02d0d5d1c Author: Eric Junyuan XieAuthorDate: Sun Nov 19 21:55:59 2017 -0800 Remove experimental warning on Gluon and add Gluon tutorials (#8710) * Remove experimental warning on Gluon and add Gluon tutorials * fix * fix * Update autograd.md --- docs/api/python/autograd/autograd.md | 5 - docs/api/python/gluon/data.md| 4 docs/api/python/gluon/gluon.md | 4 docs/api/python/gluon/model_zoo.md | 4 docs/tutorials/index.md | 34 +- 5 files changed, 29 insertions(+), 22 deletions(-) diff --git a/docs/api/python/autograd/autograd.md b/docs/api/python/autograd/autograd.md index de81884..410d6a9 100644 --- a/docs/api/python/autograd/autograd.md +++ b/docs/api/python/autograd/autograd.md @@ -1,14 +1,9 @@ # Autograd Package - ```eval_rst .. currentmodule:: mxnet.autograd ``` -```eval_rst -.. warning:: This package is currently experimental and may change in the near future. -``` - ## Overview The `autograd` package enables automatic diff --git a/docs/api/python/gluon/data.md b/docs/api/python/gluon/data.md index f72f3cd..0b5f959 100644 --- a/docs/api/python/gluon/data.md +++ b/docs/api/python/gluon/data.md @@ -15,10 +15,6 @@ This document lists the data APIs in Gluon: The `Gluon Data` API, defined in the `gluon.data` package, provides useful dataset loading and processing tools, as well as common public datasets. -```eval_rst -.. warning:: This package contains experimental APIs and may change in the near future. -``` - In the rest of this document, we list routines provided by the `gluon.data` package. ## Data diff --git a/docs/api/python/gluon/gluon.md b/docs/api/python/gluon/gluon.md index 0ef6dbe..2ae766f 100644 --- a/docs/api/python/gluon/gluon.md +++ b/docs/api/python/gluon/gluon.md @@ -5,10 +5,6 @@ .. currentmodule:: mxnet.gluon ``` -```eval_rst -.. warning:: This package is currently experimental and may change in the near future. -``` - ## Overview diff --git a/docs/api/python/gluon/model_zoo.md b/docs/api/python/gluon/model_zoo.md index 18d9ae2..a22b437 100644 --- a/docs/api/python/gluon/model_zoo.md +++ b/docs/api/python/gluon/model_zoo.md @@ -18,10 +18,6 @@ This document lists the model APIs in Gluon: The `Gluon Model Zoo` API, defined in the `gluon.model_zoo` package, provides pre-defined and pre-trained models to help bootstrap machine learning applications. -```eval_rst -.. warning:: This package contains experimental APIs and may change in the near future. -``` - In the rest of this document, we list routines provided by the `gluon.model_zoo` package. ### Vision diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 40c..6429dfb 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -2,13 +2,37 @@ These tutorials introduce a few fundamental concepts in deep learning and how to implement them in _MXNet_. The _Basics_ section contains tutorials on manipulating arrays, building networks, loading/preprocessing data, etc. The _Training and Inference_ section talks about implementing Linear Regression, training a Handwritten digit classifier using MLP and CNN, running inferences using a pre-trained model, and lastly, efficiently training a large scale image classifier. -```eval_rst -.. Note:: We are working on a set of tutorials for the new imperative interface called Gluon. A preview version is hosted at http://gluon.mxnet.io. -``` -## Python +## Gluon + +Gluon is the high-level interface for MXNet. It is more intuitive and easier to use than the lower level interface. +Gluon supports dynamic (define-by-run) graphs with JIT-compilation to achieve both flexibility and efficiency. +This is a selected subset of Gluon tutorials. For the comprehensive tutorial on Gluon, +please see [gluon.mxnet.io](http://gluon.mxnet.io). + +### Basics + +- [Manipulate data the MXNet way with ndarray](http://gluon.mxnet.io/chapter01_crashcourse/ndarray.html) +- [Automatic differentiation with autograd](http://gluon.mxnet.io/chapter01_crashcourse/autograd.html) +- [Linear regression with gluon](http://gluon.mxnet.io/chapter02_supervised-learning/linear-regression-gluon.html) +- [Serialization - saving, loading and checkpointing](http://gluon.mxnet.io/chapter03_deep-neural-networks/serialization.html) + +### Neural Networks + +- [Multilayer perceptrons in gluon](http://gluon.mxnet.io/chapter03_deep-neural-networks/mlp-gluon.html) +- [Convolutional Neural Networks in
[GitHub] liuzhi136 commented on issue #8720: Implementation Help!!!
liuzhi136 commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345598672 @szha I do not understand the meaning of "isolate the problem". Can you explain it? And Is it OK for you that I have your Wechat or email? 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: fix group2ctx with null reqs (#8717)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 7fc0396 fix group2ctx with null reqs (#8717) 7fc0396 is described below commit 7fc039639b288f80fa7fe6482de1a25e04261e5e Author: Haibin LinAuthorDate: Sun Nov 19 21:55:31 2017 -0800 fix group2ctx with null reqs (#8717) --- src/executor/graph_executor.cc | 18 + tests/python/unittest/test_multi_device_exec.py | 26 + 2 files changed, 32 insertions(+), 12 deletions(-) diff --git a/src/executor/graph_executor.cc b/src/executor/graph_executor.cc index ade8e83..01484da 100644 --- a/src/executor/graph_executor.cc +++ b/src/executor/graph_executor.cc @@ -321,6 +321,7 @@ Graph AssignContext(Graph g, const std::vector& in_arg_ctxes, const std::vector& arg_grad_ctxes, const std::vector& aux_state_ctxes, +const std::vector& grad_req_types, size_t num_forward_inputs, size_t num_forward_outputs) { const auto& idx = g.indexed_graph(); @@ -385,9 +386,15 @@ Graph AssignContext(Graph g, // loop through backward input nodes and populate maps and lists // the backward input nodes is the gradient of the loss wrt the output - for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i) { + size_t arg_grad_offset = 0; + // keep an offset into the arg_grad_ctxes vector, + // since g.outputs exclude arg_grad whose req == null + CHECK_GE(grad_req_types.size(), g.outputs.size() - num_forward_outputs) + << "insufficient number of grad_reqs"; + for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i, ++arg_grad_offset) { +while (grad_req_types[arg_grad_offset] == kNullOp) ++arg_grad_offset; const uint32_t nid = idx.outputs()[i].node_id; -Context ctx = arg_grad_ctxes[i - num_forward_outputs]; +Context ctx = arg_grad_ctxes[arg_grad_offset]; if (ctx2id.count(ctx) == 0) { ctx2id[ctx] = static_cast(ctx_list.size()); ctx_list.push_back(ctx); @@ -417,9 +424,11 @@ Graph AssignContext(Graph g, // if the assigned device of gradient node // corresponds to storage of grads auto _idx = g.indexed_graph(); - for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i) { + arg_grad_offset = 0; + for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i, ++arg_grad_offset) { +while (grad_req_types[arg_grad_offset] == kNullOp) ++arg_grad_offset; const uint32_t nid = new_idx.outputs()[i].node_id; -Context ctx = arg_grad_ctxes[i - num_forward_outputs]; +Context ctx = arg_grad_ctxes[arg_grad_offset]; CHECK(ctx == vcontext[nid]) << "Trying to save gradient to " << ctx << " while its source node \"" << new_idx[nid].source->attrs.name @@ -1055,6 +1064,7 @@ Graph GraphExecutor::InitGraph(nnvm::Symbol symbol, in_arg_ctxes, arg_grad_ctxes, aux_state_ctxes, +grad_req_types, num_forward_inputs_, num_forward_outputs_); diff --git a/tests/python/unittest/test_multi_device_exec.py b/tests/python/unittest/test_multi_device_exec.py index 0a2739d..aa279b1 100644 --- a/tests/python/unittest/test_multi_device_exec.py +++ b/tests/python/unittest/test_multi_device_exec.py @@ -20,6 +20,17 @@ import numpy as np import mxnet as mx def test_ctx_group(): +def check_ctx_group(group2ctx, grad_req, mlp, set_stage1): +texec = mlp.simple_bind(mx.cpu(0), +group2ctx=group2ctx, +data=(1,200), grad_req=grad_req) + +for arr, name in zip(texec.arg_arrays, mlp.list_arguments()): +if name in set_stage1: +assert arr.context == group2ctx['stage1'] +else: +assert arr.context == group2ctx['stage2'] + with mx.AttrScope(ctx_group='stage1'): data = mx.symbol.Variable('data') fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128) @@ -40,15 +51,14 @@ def test_ctx_group(): 'stage2' : mx.cpu(2) } -texec = mlp.simple_bind(mx.cpu(0), -group2ctx=group2ctx, -data=(1,200)) +# generate reqs with null +grad_req_with_null = {} +for arg in mlp.list_arguments(): +grad_req_with_null[arg] = 'null' if arg == 'data' else 'write' -for arr, name in zip(texec.arg_arrays, mlp.list_arguments()): -if name in set_stage1: -assert arr.context == group2ctx['stage1'] -else: -assert arr.context == group2ctx['stage2'] +grad_reqs = ['write', grad_req_with_null] +for grad_req
[GitHub] piiswrong closed pull request #8717: fix group2ctx with null reqs
piiswrong closed pull request #8717: fix group2ctx with null reqs URL: https://github.com/apache/incubator-mxnet/pull/8717 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/src/executor/graph_executor.cc b/src/executor/graph_executor.cc index ade8e838fd..01484dac29 100644 --- a/src/executor/graph_executor.cc +++ b/src/executor/graph_executor.cc @@ -321,6 +321,7 @@ Graph AssignContext(Graph g, const std::vector& in_arg_ctxes, const std::vector& arg_grad_ctxes, const std::vector& aux_state_ctxes, +const std::vector& grad_req_types, size_t num_forward_inputs, size_t num_forward_outputs) { const auto& idx = g.indexed_graph(); @@ -385,9 +386,15 @@ Graph AssignContext(Graph g, // loop through backward input nodes and populate maps and lists // the backward input nodes is the gradient of the loss wrt the output - for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i) { + size_t arg_grad_offset = 0; + // keep an offset into the arg_grad_ctxes vector, + // since g.outputs exclude arg_grad whose req == null + CHECK_GE(grad_req_types.size(), g.outputs.size() - num_forward_outputs) + << "insufficient number of grad_reqs"; + for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i, ++arg_grad_offset) { +while (grad_req_types[arg_grad_offset] == kNullOp) ++arg_grad_offset; const uint32_t nid = idx.outputs()[i].node_id; -Context ctx = arg_grad_ctxes[i - num_forward_outputs]; +Context ctx = arg_grad_ctxes[arg_grad_offset]; if (ctx2id.count(ctx) == 0) { ctx2id[ctx] = static_cast(ctx_list.size()); ctx_list.push_back(ctx); @@ -417,9 +424,11 @@ Graph AssignContext(Graph g, // if the assigned device of gradient node // corresponds to storage of grads auto _idx = g.indexed_graph(); - for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i) { + arg_grad_offset = 0; + for (size_t i = num_forward_outputs; i < g.outputs.size(); ++i, ++arg_grad_offset) { +while (grad_req_types[arg_grad_offset] == kNullOp) ++arg_grad_offset; const uint32_t nid = new_idx.outputs()[i].node_id; -Context ctx = arg_grad_ctxes[i - num_forward_outputs]; +Context ctx = arg_grad_ctxes[arg_grad_offset]; CHECK(ctx == vcontext[nid]) << "Trying to save gradient to " << ctx << " while its source node \"" << new_idx[nid].source->attrs.name @@ -1055,6 +1064,7 @@ Graph GraphExecutor::InitGraph(nnvm::Symbol symbol, in_arg_ctxes, arg_grad_ctxes, aux_state_ctxes, +grad_req_types, num_forward_inputs_, num_forward_outputs_); diff --git a/tests/python/unittest/test_multi_device_exec.py b/tests/python/unittest/test_multi_device_exec.py index 0a2739d9bb..aa279b1837 100644 --- a/tests/python/unittest/test_multi_device_exec.py +++ b/tests/python/unittest/test_multi_device_exec.py @@ -20,6 +20,17 @@ import mxnet as mx def test_ctx_group(): +def check_ctx_group(group2ctx, grad_req, mlp, set_stage1): +texec = mlp.simple_bind(mx.cpu(0), +group2ctx=group2ctx, +data=(1,200), grad_req=grad_req) + +for arr, name in zip(texec.arg_arrays, mlp.list_arguments()): +if name in set_stage1: +assert arr.context == group2ctx['stage1'] +else: +assert arr.context == group2ctx['stage2'] + with mx.AttrScope(ctx_group='stage1'): data = mx.symbol.Variable('data') fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128) @@ -40,15 +51,14 @@ def test_ctx_group(): 'stage2' : mx.cpu(2) } -texec = mlp.simple_bind(mx.cpu(0), -group2ctx=group2ctx, -data=(1,200)) +# generate reqs with null +grad_req_with_null = {} +for arg in mlp.list_arguments(): +grad_req_with_null[arg] = 'null' if arg == 'data' else 'write' -for arr, name in zip(texec.arg_arrays, mlp.list_arguments()): -if name in set_stage1: -assert arr.context == group2ctx['stage1'] -else: -assert arr.context == group2ctx['stage2'] +grad_reqs = ['write', grad_req_with_null] +for grad_req in grad_reqs: +check_ctx_group(group2ctx, grad_req, mlp, set_stage1) def test_ctx_group_sparse(): with mx.AttrScope(ctx_group='stage1'): This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub
[GitHub] leezu commented on issue #7931: MKL-DNN integration: request for reviews
leezu commented on issue #7931: MKL-DNN integration: request for reviews URL: https://github.com/apache/incubator-mxnet/pull/7931#issuecomment-345598574 @sbodenstein MKL with v0.11 is quite buggy. I often got `inf` values during training for no obvious reason (i.e. training is stable without mkl / on GPU). I'm not sure about v0.12 due to https://github.com/apache/incubator-mxnet/issues/8280 . (i.e. even without MKL-DNN MKL is buggy) 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 With regards, Apache Git Services
[GitHub] leezu commented on issue #7931: MKL-DNN integration: request for reviews
leezu commented on issue #7931: MKL-DNN integration: request for reviews URL: https://github.com/apache/incubator-mxnet/pull/7931#issuecomment-345598574 @sbodenstein MKL with v0.11 is quite buggy. I often got `inf` values during training for no obvious reason (i.e. training is stable without mkl / on GPU). I'm not sure about v0.12 due to https://github.com/apache/incubator-mxnet/issues/8280 . 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 With regards, Apache Git Services
[GitHub] Jerryzcn closed issue #8723: nn.block.save_params not working properly.
Jerryzcn closed issue #8723: nn.block.save_params not working properly. URL: https://github.com/apache/incubator-mxnet/issues/8723 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 With regards, Apache Git Services
[GitHub] Jerryzcn commented on issue #8723: nn.block.save_params not working properly.
Jerryzcn commented on issue #8723: nn.block.save_params not working properly. URL: https://github.com/apache/incubator-mxnet/issues/8723#issuecomment-345597125 I see thanks! Closing issue. 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8720: Implementation Help!!!
szha commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345595431 What have you tried to isolate the problem? 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8723: nn.block.save_params not working properly.
szha commented on issue #8723: nn.block.save_params not working properly. URL: https://github.com/apache/incubator-mxnet/issues/8723#issuecomment-345593756 I see. block.save_params has the option to strip away the network prefix which is on by default. turning it off should fix the problem. Alternatively you can explicitly set the prefix of a block to be empty. 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 With regards, Apache Git Services
[GitHub] anirudh2290 commented on a change in pull request #8721: fix custom op for backward compatibility
anirudh2290 commented on a change in pull request #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721#discussion_r151902607 ## File path: tests/python/unittest/test_operator.py ## @@ -3652,6 +3652,42 @@ def create_operator(self, ctx, shapes, dtypes): assert (x.grad.stype == 'csr') assert (y.stype == 'csr') assert (aux.stype == 'csr') + +# test for backward compatibility, i.e. the correctness of default implementation of Review comment: Please remove whitespaces here and above. 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 With regards, Apache Git Services
[GitHub] anirudh2290 commented on issue #8721: fix custom op for backward compatibility
anirudh2290 commented on issue #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721#issuecomment-345591574 Thank you for fixing this and adding the tests! Yes, since the output of stypes which is the in_grads would depend on arguments length and not output length. Can you please let me know what LR example you are referring to? 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 With regards, Apache Git Services
[GitHub] eric-haibin-lin commented on issue #8721: fix custom op for backward compatibility
eric-haibin-lin commented on issue #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721#issuecomment-345590950 @anirudh2290 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 With regards, Apache Git Services
[GitHub] mbaijal commented on issue #8175: WIP: Julia CI build
mbaijal commented on issue #8175: WIP: Julia CI build URL: https://github.com/apache/incubator-mxnet/pull/8175#issuecomment-345590057 Hi @iblis17 I should be able to help you with this. You can ping me on slack at mbaijal. 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 With regards, Apache Git Services
[GitHub] Jerryzcn commented on issue #8723: nn.block.save_params not working properly.
Jerryzcn commented on issue #8723: nn.block.save_params not working properly. URL: https://github.com/apache/incubator-mxnet/issues/8723#issuecomment-345589651 AssertionError: Parameter net0_conv0_weight is missing in file checkpoint\latest-net-0-0.0.params. Basically, the weights are saved using different name then the load function. 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 With regards, Apache Git Services
[GitHub] zhreshold commented on issue #8716: ImageDetIter uses type list as inside making as_in_context break
zhreshold commented on issue #8716: ImageDetIter uses type list as inside making as_in_context break URL: https://github.com/apache/incubator-mxnet/issues/8716#issuecomment-345589253 `as_in_context` don't support list of context. 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 With regards, Apache Git Services
[GitHub] zhreshold commented on issue #8715: Documentation for MultiBoxPrior is not substantial
zhreshold commented on issue #8715: Documentation for MultiBoxPrior is not substantial URL: https://github.com/apache/incubator-mxnet/issues/8715#issuecomment-345589157 You can try help(mx.nd.contrib.MultiBoxPrior) for help info in python. However, I have no idea why the generated function signature are really bad. @szha Any clue? 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 With regards, Apache Git Services
[GitHub] jiajiechen commented on issue #8684: [BugFix][CoreML Converter] Dense layers w/o bias.
jiajiechen commented on issue #8684: [BugFix][CoreML Converter] Dense layers w/o bias. URL: https://github.com/apache/incubator-mxnet/pull/8684#issuecomment-345588252 This looks good to me. nitpick: I will suggest ```python has_bias = not param.pop('no_bias', False) ``` 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 With regards, Apache Git Services
[GitHub] szha commented on issue #8723: nn.block.save_params not working properly.
szha commented on issue #8723: nn.block.save_params not working properly. URL: https://github.com/apache/incubator-mxnet/issues/8723#issuecomment-345587528 It should work both way. What's the error that you saw? 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 With regards, Apache Git Services
[GitHub] liuzhi136 commented on issue #8720: Implementation Help!!!
liuzhi136 commented on issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720#issuecomment-345585456 @mli @piiswrong @szha 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 With regards, Apache Git Services
[GitHub] iblis17 commented on issue #8175: WIP: Julia CI build
iblis17 commented on issue #8175: WIP: Julia CI build URL: https://github.com/apache/incubator-mxnet/pull/8175#issuecomment-345583668 bump? 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 With regards, Apache Git Services
[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict
liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-345583241 @eric-haibin-lin ok,i will try later 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 With regards, Apache Git Services
[GitHub] YujiOshima opened a new pull request #8722: Profiler: set cpu/gpu num during execution
YujiOshima opened a new pull request #8722: Profiler: set cpu/gpu num during execution URL: https://github.com/apache/incubator-mxnet/pull/8722 Signed-off-by: YujiOshima## Description ## In Profiler, set CPU and GPU number dynamically. Enable to use profiler in a large number of CPU or GPU environment. @ZihengJiang ## Checklist ## ### Essentials ### - [ ] Passed code style checking (`make lint`) - [ ] Changes are complete (i.e. I finished coding on this PR) - [ ] All changes have test coverage - [ ] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. - [ ] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here 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 With regards, Apache Git Services
[GitHub] atticcas commented on issue #8308: Error with Python custom operator in distributed training
atticcas commented on issue #8308: Error with Python custom operator in distributed training URL: https://github.com/apache/incubator-mxnet/issues/8308#issuecomment-345576806 Ran in to the same problem. Did some dive into the code and it seems like something with how MXNet handles distributed computation. 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 With regards, Apache Git Services
[GitHub] ZiyueHuang commented on issue #8500: program crash when run sparse model predict
ZiyueHuang commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-345576013 Current LR example is broken due to the custom op. This is fixed in https://github.com/apache/incubator-mxnet/pull/8721. 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 With regards, Apache Git Services
[GitHub] ZiyueHuang opened a new pull request #8721: fix custom op for backward compatibility
ZiyueHuang opened a new pull request #8721: fix custom op for backward compatibility URL: https://github.com/apache/incubator-mxnet/pull/8721 ## Description ## Before this PR, the LR in sparse example is broken due to to the custom op, and the added test will throw err message, ``` Error in mult.infer_type: Traceback (most recent call last): File "/home/hanfeng/zyh/mxnet/python/mxnet/operator.py", line 737, in infer_storage_type_backward_entry "stypes, got %d."%(total_outputs, len(ostype)) AssertionError: InferStorageTypeBackward Error: expecting 2 entries in returned output stypes, got 1. [10:27:12] /home/hanfeng/zyh/mxnet/dmlc-core/include/dmlc/./logging.h:308: [10:27:12] src/operator/custom/custom.cc:383: Check failed: reinterpret_cast( params.info->callbacks[kCustomOpPropBackwardInferStorageType])( stypes.size(), stypes.data(), params.info->contexts[kCustomOpPropBackwardInferStorageType]) ``` cc @eric-haibin-lin @anirudh2290 ## Checklist ## ### Essentials ### - [ ] Passed code style checking (`make lint`) - [ ] Changes are complete (i.e. I finished coding on this PR) - [ ] All changes have test coverage - [ ] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. - [ ] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here 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 With regards, Apache Git Services
[GitHub] solin319 commented on issue #8107: "add warmup lr_scheduler" create a new pr
solin319 commented on issue #8107: "add warmup lr_scheduler" create a new pr URL: https://github.com/apache/incubator-mxnet/pull/8107#issuecomment-345572425 I have update the parameters name and description. @piiswrong 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 With regards, Apache Git Services
[GitHub] yajiedesign commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2?
yajiedesign commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2? URL: https://github.com/apache/incubator-mxnet/issues/8703#issuecomment-345569605 synchronization with numpy,stop add new feature when 2019.1.1 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 With regards, Apache Git Services
[GitHub] yajiedesign commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2?
yajiedesign commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2? URL: https://github.com/apache/incubator-mxnet/issues/8703#issuecomment-345569605 like numpy,stop add new feature when 2019.1.1 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 With regards, Apache Git Services
[GitHub] yajiedesign commented on issue #8714: win10 cuda8.0 vs2013 compile problem
yajiedesign commented on issue #8714: win10 cuda8.0 vs2013 compile problem URL: https://github.com/apache/incubator-mxnet/issues/8714#issuecomment-345569063 include submodule? try use vs2015. 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 With regards, Apache Git Services
[GitHub] yajiedesign commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
yajiedesign commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345568868 @cjolivier01 yes, lint is out of date,but fix it not difficulty . 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 With regards, Apache Git Services
[GitHub] liuzhi136 opened a new issue #8720: Implementation Help!!!
liuzhi136 opened a new issue #8720: Implementation Help!!! URL: https://github.com/apache/incubator-mxnet/issues/8720 I'm trying to implement SAN model proposed in "Stacked Attention Network for Image Question Answering" in mxnet, which can be download below. [Yang_Stacked_Attention_Networks_CVPR_2016_paper.pdf](https://github.com/apache/incubator-mxnet/files/1486170/Yang_Stacked_Attention_Networks_CVPR_2016_paper.pdf) My implementations looks like below : [san_vqamodel_with_gluon.txt](https://github.com/apache/incubator-mxnet/files/1486193/san_vqamodel_with_gluon.txt) I had downloaded the DAQUAR dataset and use the reduced version to reproduce the result of this paper. I think the model described clearly enough in this paper. However, this model I implement doesn't converge at all and always fluctuate after several epochs, and its training and validation error less than 3. I've tried various hyperparameter sets to adjust the training procedure and does not work out. It's urgent problem for me, and I really need help to solve it as soon as possible. I really want to know if my implementation is correct. Does any could help me? Cause it's a really urgent. Any help will be appreciated! 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 With regards, Apache Git Services
[GitHub] BiranLi closed issue #7665: Request on supporting Pooling symbol with NHWC.
BiranLi closed issue #7665: Request on supporting Pooling symbol with NHWC. URL: https://github.com/apache/incubator-mxnet/issues/7665 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345564863 I expect that there are many users of MXNet finding implementation of CapsNet. We're in the middle of implementing data augmentation and will make effort to follow up the accuracy soon. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345564863 I expect that there are many users of MXNet finding implementation of CapsNet. We're in the middle of implement data augmentation and will make effort to follow up the accuracy soon. 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 With regards, Apache Git Services
[GitHub] burness commented on issue #8621: fix row_sparse demo tutorials doc
burness commented on issue #8621: fix row_sparse demo tutorials doc URL: https://github.com/apache/incubator-mxnet/pull/8621#issuecomment-345564109 Ok! I will try @eric-haibin-lin 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 With regards, Apache Git Services
[GitHub] cjolivier01 opened a new pull request #8719: POC Tune without Launch specialization macros (WIP for post 1.0 rc0)
cjolivier01 opened a new pull request #8719: POC Tune without Launch specialization macros (WIP for post 1.0 rc0) URL: https://github.com/apache/incubator-mxnet/pull/8719 @piiswrong ## Description ## (Brief description on what this PR is about) ## Checklist ## ### Essentials ### - [ ] Passed code style checking (`make lint`) - [ ] Changes are complete (i.e. I finished coding on this PR) - [ ] All changes have test coverage - [ ] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. - [ ] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: Revert "2bit gradient compression (#8662)" (#8711)
This is an automated email from the ASF dual-hosted git repository. zhasheng pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 504f42f Revert "2bit gradient compression (#8662)" (#8711) 504f42f is described below commit 504f42ffc5ae26980d39da199516c37d7f9fc56c Author: Sheng ZhaAuthorDate: Sun Nov 19 15:51:42 2017 -0800 Revert "2bit gradient compression (#8662)" (#8711) This reverts commit a499f892c9ee6f59ccfb57c9e431c91014078891. --- example/image-classification/common/fit.py | 44 ++-- example/rnn/lstm_bucketing.py | 1 + include/mxnet/c_api.h | 13 - include/mxnet/kvstore.h| 15 -- python/mxnet/gluon/trainer.py | 12 +- python/mxnet/kvstore.py| 62 - python/mxnet/module/bucketing_module.py| 17 +- python/mxnet/module/module.py | 11 +- src/c_api/c_api.cc | 14 -- src/kvstore/comm.h | 87 +-- src/kvstore/gradient_compression-inl.h | 155 src/kvstore/gradient_compression.cc| 193 -- src/kvstore/gradient_compression.cu| 40 --- src/kvstore/gradient_compression.h | 138 -- src/kvstore/kvstore.cc | 2 +- src/kvstore/kvstore_dist.h | 388 - src/kvstore/kvstore_dist_server.h | 143 ++- src/kvstore/kvstore_local.h| 7 - tests/nightly/dist_sync_kvstore.py | 120 + tests/nightly/test_kvstore.py | 200 ++- tools/bandwidth/measure.py | 6 +- 21 files changed, 167 insertions(+), 1501 deletions(-) diff --git a/example/image-classification/common/fit.py b/example/image-classification/common/fit.py index 2b002c7..51a1abe 100755 --- a/example/image-classification/common/fit.py +++ b/example/image-classification/common/fit.py @@ -103,11 +103,6 @@ def add_fit_args(parser): help='1 means test reading speed without training') train.add_argument('--dtype', type=str, default='float32', help='precision: float32 or float16') -train.add_argument('--gc-type', type=str, default='none', - help='type of gradient compression to use, \ - takes `2bit` or `none` for now') -train.add_argument('--gc-threshold', type=float, default=0.5, - help='threshold for 2bit gradient compression') return train def fit(args, network, data_loader, **kwargs): @@ -119,9 +114,6 @@ def fit(args, network, data_loader, **kwargs): """ # kvstore kv = mx.kvstore.create(args.kv_store) -if args.gc_type != 'none': -kv.set_gradient_compression({'type': args.gc_type, - 'threshold': args.gc_threshold}) # logging head = '%(asctime)-15s Node[' + str(kv.rank) + '] %(message)s' @@ -170,10 +162,10 @@ def fit(args, network, data_loader, **kwargs): lr_scheduler = lr_scheduler optimizer_params = { -'learning_rate': lr, -'wd' : args.wd, -'lr_scheduler': lr_scheduler, -'multi_precision': True} +'learning_rate': lr, +'wd' : args.wd, +'lr_scheduler': lr_scheduler, +'multi_precision': True} # Only a limited number of optimizers have 'momentum' property has_momentum = {'sgd', 'dcasgd', 'nag'} @@ -203,17 +195,17 @@ def fit(args, network, data_loader, **kwargs): # run model.fit(train, - begin_epoch= args.load_epoch if args.load_epoch else 0, - num_epoch = args.num_epochs, - eval_data = val, - eval_metric= eval_metrics, - kvstore= kv, - optimizer = args.optimizer, - optimizer_params = optimizer_params, - initializer= initializer, - arg_params = arg_params, - aux_params = aux_params, - batch_end_callback = batch_end_callbacks, - epoch_end_callback = checkpoint, - allow_missing = True, - monitor= monitor) +begin_epoch= args.load_epoch if args.load_epoch else 0, +num_epoch = args.num_epochs, +eval_data = val, +eval_metric= eval_metrics, +kvstore= kv, +optimizer = args.optimizer, +optimizer_params = optimizer_params, +initializer= initializer, +arg_params = arg_params, +aux_params = aux_params, +batch_end_callback = batch_end_callbacks, +
[GitHub] szha closed pull request #8711: Revert "2bit gradient compression"
szha closed pull request #8711: Revert "2bit gradient compression" URL: https://github.com/apache/incubator-mxnet/pull/8711 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 With regards, Apache Git Services
[GitHub] eric-haibin-lin commented on issue #8707: Fail to build amalgamation for Android in latest version
eric-haibin-lin commented on issue #8707: Fail to build amalgamation for Android in latest version URL: https://github.com/apache/incubator-mxnet/issues/8707#issuecomment-345557830 @arank could you help diagnose the issue? 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 With regards, Apache Git Services
[GitHub] eric-haibin-lin commented on issue #8500: program crash when run sparse model predict
eric-haibin-lin commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-345557450 The cpu _backward_dot operator was improved by at least 3x in #8611 Do you want to sync with master and run it again? 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.25% and we expect close accuracy(0.34%) can be achieved by adding data augmentation(scale shift) like other implementations of other platforms(keras, tensorflow) which achieved 0.34 % (the average) result. The result of 0.25 is hard to achieve but we expect close enough accuracy can be obtained soon. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.25% and we expect it can be achieved by adding data augmentation(scale shift) like other implementations of other platforms(keras, tensorflow) which achieved 0.34 % (the average) result. The result of 0.25 is hard to achieve but we expect close enough accuracy can be obtained soon. 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 With regards, Apache Git Services
[GitHub] mbaijal commented on issue #8401: Make make lint compatible with python3 (don't call python2 explicitly)
mbaijal commented on issue #8401: Make make lint compatible with python3 (don't call python2 explicitly) URL: https://github.com/apache/incubator-mxnet/pull/8401#issuecomment-34127 Hi @larroy I think you need to rebase this PR with latest master to make it pass unit tests. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.25% and we expect it can be achieved by adding data augmentation(scale shift) like other implementations of other platforms(keras, tensorflow) which achieved 0.34 % (the average) result. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.25% and we expect it can be achieved by adding data augmentation(scale shift) like other implementations of other platforms(keras, tensorflow) which achieved the same result of paper. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.39% and we expect it can be achieved by adding data augmentation(scale shift) like other implementations of other platforms(keras, tensorflow) which achieved the same result of paper. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.39% and we expect it can be achieved by adding data augmentation(scale shift) like other implementations of other platforms which achieved the same result of paper. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on issue #8674: ADD CapsNet example
Soonhwan-Kwon commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345554566 The original paper's result was 0.39% and we expect it can be achieved by adding data augmentation(scale shift) like other implementation of other platforms. 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 With regards, Apache Git Services
[GitHub] Soonhwan-Kwon commented on a change in pull request #8674: ADD CapsNet example
Soonhwan-Kwon commented on a change in pull request #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#discussion_r151879998 ## File path: example/capsnet/README.md ## @@ -0,0 +1,32 @@ +**CapsNet-MXNet** += + +This example is MXNet implementation of [CapsNet](https://arxiv.org/abs/1710.09829): +Sara Sabour, Nicholas Frosst, Geoffrey E Hinton. Dynamic Routing Between Capsules. NIPS 2017 +- The Current best test error is 0.5% + +Due to the permission issue, this example is maintained in this [repository](https://github.com/samsungsds-rnd/capsnet.mxnet) separately. Review comment: If you mean the last line, Pull Request can takes days to accept, meanwhile some follow-up changes can be updated to our temporary repository https://github.com/samsungsds-rnd/capsnet.mxnet. 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 With regards, Apache Git Services
[GitHub] szha closed issue #8319: [WIP-NewFeature] ONNX support for MXNet
szha closed issue #8319: [WIP-NewFeature] ONNX support for MXNet URL: https://github.com/apache/incubator-mxnet/issues/8319 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 With regards, Apache Git Services
[GitHub] szha opened a new pull request #8718: Doc src and fix
szha opened a new pull request #8718: Doc src and fix URL: https://github.com/apache/incubator-mxnet/pull/8718 ## Description ## Add a link for viewing source code in api doc. Updated doc can be found at http://mxnet-doc.s3-accelerate.dualstack.amazonaws.com/api/python/index.html ## Checklist ## ### Essentials ### - [x] Passed code style checking (`make lint`) - [x] Changes are complete (i.e. I finished coding on this PR) - [x] All changes have test coverage - [x] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. - [x] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [x] add `sphinx.ext.viewcode` - [x] fix links ## Comments ## - Code that is generated at run-time isn't available for viewing (e.g. frontend functions for operators) 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 With regards, Apache Git Services
[GitHub] piiswrong commented on a change in pull request #8107: "add warmup lr_scheduler" create a new pr
piiswrong commented on a change in pull request #8107: "add warmup lr_scheduler" create a new pr URL: https://github.com/apache/incubator-mxnet/pull/8107#discussion_r151878202 ## File path: python/mxnet/lr_scheduler.py ## @@ -100,16 +100,27 @@ class MultiFactorScheduler(LRScheduler): Then calculate the new learning rate by:: - base_lr * pow(factor, k+1) +base_lr * pow(factor, k+1) + +When warmup_step>1, warmup the learning rate by a const value for first warmup_step steps. +It returns a new learning rate by:: + +begin_lr + (num_update - 1) * const_update Parameters -- step: list of int -The list of steps to schedule a change +The list of steps to schedule a change. factor: float The factor to change the learning rate. +warmup_step : int +Changes the learning rate for first 'warmup_step' updates. +begin_lr : float, optional +The learning rate at begin. +stop_lr : float, optional +Stop updating the learning rate if it is less than this value. Review comment: begin_lr and stop_lr are confusing names and the does doesn't explain exactly what they mean 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: replace `has_key` by `in` (#8317)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 7f7e13d replace `has_key` by `in` (#8317) 7f7e13d is described below commit 7f7e13dee74ceda384b6201e71beccb1e71a8c31 Author: Chih-Ming ChenAuthorDate: Mon Nov 20 05:07:48 2017 +0800 replace `has_key` by `in` (#8317) * replace `has_key` by `in` * Update utils.py --- python/mxnet/notebook/callback.py | 2 +- tests/nightly/TestDoc/doc_spell_checker.py | 2 +- tools/accnn/rank_selection.py | 2 +- tools/accnn/utils.py | 5 +++-- 4 files changed, 6 insertions(+), 5 deletions(-) diff --git a/python/mxnet/notebook/callback.py b/python/mxnet/notebook/callback.py index 56321b7..776900f 100644 --- a/python/mxnet/notebook/callback.py +++ b/python/mxnet/notebook/callback.py @@ -367,7 +367,7 @@ class LiveLearningCurve(LiveBokehChart): metrics = {} metrics['elapsed'] = datetime.datetime.now() - self.start_time for key, value in metrics.items(): -if not self._data[df_name].has_key(key): +if key not in self._data[df_name]: self._data[df_name][key] = [] self._data[df_name][key].append(value) diff --git a/tests/nightly/TestDoc/doc_spell_checker.py b/tests/nightly/TestDoc/doc_spell_checker.py index a7b8b25..a33807e 100644 --- a/tests/nightly/TestDoc/doc_spell_checker.py +++ b/tests/nightly/TestDoc/doc_spell_checker.py @@ -92,7 +92,7 @@ def check_doc(file_content, spell_checker, spell_check_ret): """ spell_checker.set_text(file_content) for error in spell_checker: -if spell_check_ret.has_key(error.word): +if error.word in spell_check_ret: spell_check_ret[error.word] += 1 else: spell_check_ret[error.word] = 1 diff --git a/tools/accnn/rank_selection.py b/tools/accnn/rank_selection.py index 66937b2..c5c0261 100644 --- a/tools/accnn/rank_selection.py +++ b/tools/accnn/rank_selection.py @@ -81,7 +81,7 @@ def get_ranksel(model, ratio): if nxt_c > EC: continue nxt_v = dp[now][now_c] + math.log(S[i][d]) -if dp[nxt].has_key(nxt_c): +if nxt_c in dp[nxt]: if nxt_v > dp[nxt][nxt_c]: dp[nxt][nxt_c] = nxt_v dpc[i][nxt_c] = (d,now_c) diff --git a/tools/accnn/utils.py b/tools/accnn/utils.py index 25fb188..2795f85 100644 --- a/tools/accnn/utils.py +++ b/tools/accnn/utils.py @@ -20,6 +20,7 @@ import copy import json import ast + def load_model(args): devs = mx.cpu() if args.gpus == None else [mx.gpu(int(i)) for i in args.gpus.split(',')] return mx.model.FeedForward.load(args.model, args.load_epoch, ctx=devs) @@ -29,7 +30,7 @@ def topsort(nodes): deg = [0]*n g = [[] for _ in xrange(n)] for i,node in enumerate(nodes): -if node.has_key('inputs'): +if 'inputs' in node: for j in node['inputs']: deg[i] += 1 g[j[0]].append(i) @@ -45,7 +46,7 @@ def topsort(nodes): q.append(j) new_ids=dict([(node['name'],i) for i,node in enumerate(res)]) for node in res: -if node.has_key('inputs'): +if 'inputs' in node: for j in node['inputs']: j[0]=new_ids[nodes[j[0]]['name']] return res -- To stop receiving notification emails like this one, please contact ['"comm...@mxnet.apache.org" '].
[GitHub] piiswrong closed pull request #8317: replace `has_key` by `in`
piiswrong closed pull request #8317: replace `has_key` by `in` URL: https://github.com/apache/incubator-mxnet/pull/8317 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/python/mxnet/notebook/callback.py b/python/mxnet/notebook/callback.py index 56321b715b..776900fe59 100644 --- a/python/mxnet/notebook/callback.py +++ b/python/mxnet/notebook/callback.py @@ -367,7 +367,7 @@ def _process_batch(self, param, df_name): metrics = {} metrics['elapsed'] = datetime.datetime.now() - self.start_time for key, value in metrics.items(): -if not self._data[df_name].has_key(key): +if key not in self._data[df_name]: self._data[df_name][key] = [] self._data[df_name][key].append(value) diff --git a/tests/nightly/TestDoc/doc_spell_checker.py b/tests/nightly/TestDoc/doc_spell_checker.py index a7b8b250c9..a33807e3d5 100644 --- a/tests/nightly/TestDoc/doc_spell_checker.py +++ b/tests/nightly/TestDoc/doc_spell_checker.py @@ -92,7 +92,7 @@ def check_doc(file_content, spell_checker, spell_check_ret): """ spell_checker.set_text(file_content) for error in spell_checker: -if spell_check_ret.has_key(error.word): +if error.word in spell_check_ret: spell_check_ret[error.word] += 1 else: spell_check_ret[error.word] = 1 diff --git a/tools/accnn/rank_selection.py b/tools/accnn/rank_selection.py index 66937b2859..c5c026114a 100644 --- a/tools/accnn/rank_selection.py +++ b/tools/accnn/rank_selection.py @@ -81,7 +81,7 @@ def get_ranksel(model, ratio): if nxt_c > EC: continue nxt_v = dp[now][now_c] + math.log(S[i][d]) -if dp[nxt].has_key(nxt_c): +if nxt_c in dp[nxt]: if nxt_v > dp[nxt][nxt_c]: dp[nxt][nxt_c] = nxt_v dpc[i][nxt_c] = (d,now_c) diff --git a/tools/accnn/utils.py b/tools/accnn/utils.py index 25fb188956..2795f8558f 100644 --- a/tools/accnn/utils.py +++ b/tools/accnn/utils.py @@ -20,6 +20,7 @@ import json import ast + def load_model(args): devs = mx.cpu() if args.gpus == None else [mx.gpu(int(i)) for i in args.gpus.split(',')] return mx.model.FeedForward.load(args.model, args.load_epoch, ctx=devs) @@ -29,7 +30,7 @@ def topsort(nodes): deg = [0]*n g = [[] for _ in xrange(n)] for i,node in enumerate(nodes): -if node.has_key('inputs'): +if 'inputs' in node: for j in node['inputs']: deg[i] += 1 g[j[0]].append(i) @@ -45,7 +46,7 @@ def topsort(nodes): q.append(j) new_ids=dict([(node['name'],i) for i,node in enumerate(res)]) for node in res: -if node.has_key('inputs'): +if 'inputs' in node: for j in node['inputs']: j[0]=new_ids[nodes[j[0]]['name']] return res 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 With regards, Apache Git Services
[GitHub] piiswrong closed pull request #8611: optimization for dot(csr.T, dense) = rsp
piiswrong closed pull request #8611: optimization for dot(csr.T, dense) = rsp URL: https://github.com/apache/incubator-mxnet/pull/8611 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/src/operator/tensor/dot-inl.h b/src/operator/tensor/dot-inl.h index 7ab4710090..2432703291 100644 --- a/src/operator/tensor/dot-inl.h +++ b/src/operator/tensor/dot-inl.h @@ -30,9 +30,10 @@ #include #include #include -#include "./init_op.h" +#include "./util/tensor_util-inl.h" #include "../mshadow_op.h" #include "../elemwise_op_common.h" +#include "./init_op.h" #include "../mxnet_op.h" #ifdef __CUDACC__ #include "./dot-inl.cuh" @@ -364,19 +365,17 @@ struct DotCsrTransDnsDnsByRowBlocks { /*! * \brief CPU Kernel of dot(csr.T(), dns) = rsp - * Parallelization by row blocks. - * This kernel fills up the row_idx array of the rsp - * with 1 for nonzero rows and 0 for zero rows. - * The matrix will be compacted after this kernel call. + * Parallelization by row blocks which evenly partition the non-zero rows. */ struct DotCsrTransDnsRspByRowBlocks { /*! * \brief * \param i the i-th thread */ - template + template MSHADOW_CINLINE static void Map(int i, DType* out, + nnvm::dim_t* row_flg_sum, RType* row_idx, const DType* data_l, const IType* indptr_l, @@ -384,21 +383,25 @@ struct DotCsrTransDnsRspByRowBlocks { const DType* data_r, const nnvm::dim_t seg_len, const nnvm::dim_t num_rows_l, - const nnvm::dim_t num_rows, + const nnvm::dim_t nnr, const nnvm::dim_t num_cols) { using nnvm::dim_t; const dim_t seg_start = i * seg_len; -if (seg_start >= num_rows) return; +if (seg_start >= nnr) return; const dim_t seg_end = (i + 1) * seg_len; +const dim_t col_start = row_idx[seg_start]; +const dim_t col_end = seg_end >= nnr ? (row_idx[nnr-1] + 1) : row_idx[seg_end]; for (dim_t j = 0; j < num_rows_l; ++j) { if (indptr_l[j] == indptr_l[j+1]) continue; const dim_t offset_r = j * num_cols; for (IType k = indptr_l[j]; k < indptr_l[j+1]; ++k) { const CType col_idx = col_idx_l[k]; -if (col_idx < seg_start || col_idx >= seg_end) continue; -const dim_t offset_out = col_idx * num_cols; -row_idx[col_idx] = 1; +if (col_idx < col_start || col_idx >= col_end) continue; + +const nnvm::dim_t rsp_row = row_flg_sum[col_idx] - 1; +const nnvm::dim_t offset_out = rsp_row * num_cols; const DType val = data_l[k]; + for (dim_t l = 0; l < num_cols; ++l) { out[offset_out+l] += data_r[offset_r+l] * val; } @@ -605,43 +608,51 @@ inline void DotCsrDnsRspImpl(const OpContext& ctx, const TBlob col_idx_l = lhs.aux_data(csr::kIdx); const TBlob& data_r = rhs; - // pre-allocate spaces for ret using the dense dimension size - ret->CheckAndAlloc({mshadow::Shape1(lhs.shape()[1])}); - const TBlob data_out = ret->data(); - const TBlob row_idx_out = ret->aux_data(rowsparse::kIdx); - MSHADOW_SGL_DBL_TYPE_SWITCH(data_l.type_flag_, DType, { // data type MSHADOW_IDX_TYPE_SWITCH(indptr_l.type_flag_, IType, { // indptr type MSHADOW_IDX_TYPE_SWITCH(col_idx_l.type_flag_, CType, { // col idx type -MSHADOW_IDX_TYPE_SWITCH(row_idx_out.type_flag_, RType, { // row idx type +MSHADOW_IDX_TYPE_SWITCH(ret->aux_type(rowsparse::kIdx), RType, { // row idx type + const dim_t num_rows = lhs.shape()[1]; + size_t workspace_size = 2 * (num_rows * sizeof(dim_t)); + mshadow::Tensorworkspace = +ctx.requested[0].get_space_typed ( +mshadow::Shape1(workspace_size), s); + dim_t* row_flg = reinterpret_cast (workspace.dptr_); + dim_t* prefix_sum = row_flg + num_rows; + + Fill(s, TBlob(row_flg, mshadow::Shape1(num_rows), cpu::kDevMask), kWriteTo, 0); + mxnet_op::Kernel ::Launch(s, lhs.aux_shape(csr::kIdx)[0], row_flg, +col_idx_l.dptr()); + + prefix_sum[0] = row_flg[0]; + for (nnvm::dim_t i = 1; i < num_rows; i++) { +prefix_sum[i] = prefix_sum[i - 1] + row_flg[i]; + } + dim_t nnr = prefix_sum[num_rows - 1]; + + if (nnr == 0) { +FillZerosRspImpl(s, *ret); +return; + } + + ret->CheckAndAlloc({mshadow::Shape1(nnr)}); +
[incubator-mxnet] branch master updated: optimization for dot(csr.T, dense) = rsp (#8611)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new f79d22d optimization for dot(csr.T, dense) = rsp (#8611) f79d22d is described below commit f79d22db25847453a9a286eb19e9064c246a82d4 Author: Ziyue HuangAuthorDate: Mon Nov 20 05:07:10 2017 +0800 optimization for dot(csr.T, dense) = rsp (#8611) * optimization for dot(csr.T, dense) = rsp * remove unneccessary headers * load balance * minor fix and update comments * resolve * trigger * trigger --- src/operator/tensor/dot-inl.h | 93 --- 1 file changed, 52 insertions(+), 41 deletions(-) diff --git a/src/operator/tensor/dot-inl.h b/src/operator/tensor/dot-inl.h index 7ab4710..2432703 100644 --- a/src/operator/tensor/dot-inl.h +++ b/src/operator/tensor/dot-inl.h @@ -30,9 +30,10 @@ #include #include #include -#include "./init_op.h" +#include "./util/tensor_util-inl.h" #include "../mshadow_op.h" #include "../elemwise_op_common.h" +#include "./init_op.h" #include "../mxnet_op.h" #ifdef __CUDACC__ #include "./dot-inl.cuh" @@ -364,19 +365,17 @@ struct DotCsrTransDnsDnsByRowBlocks { /*! * \brief CPU Kernel of dot(csr.T(), dns) = rsp - * Parallelization by row blocks. - * This kernel fills up the row_idx array of the rsp - * with 1 for nonzero rows and 0 for zero rows. - * The matrix will be compacted after this kernel call. + * Parallelization by row blocks which evenly partition the non-zero rows. */ struct DotCsrTransDnsRspByRowBlocks { /*! * \brief * \param i the i-th thread */ - template + template MSHADOW_CINLINE static void Map(int i, DType* out, + nnvm::dim_t* row_flg_sum, RType* row_idx, const DType* data_l, const IType* indptr_l, @@ -384,21 +383,25 @@ struct DotCsrTransDnsRspByRowBlocks { const DType* data_r, const nnvm::dim_t seg_len, const nnvm::dim_t num_rows_l, - const nnvm::dim_t num_rows, + const nnvm::dim_t nnr, const nnvm::dim_t num_cols) { using nnvm::dim_t; const dim_t seg_start = i * seg_len; -if (seg_start >= num_rows) return; +if (seg_start >= nnr) return; const dim_t seg_end = (i + 1) * seg_len; +const dim_t col_start = row_idx[seg_start]; +const dim_t col_end = seg_end >= nnr ? (row_idx[nnr-1] + 1) : row_idx[seg_end]; for (dim_t j = 0; j < num_rows_l; ++j) { if (indptr_l[j] == indptr_l[j+1]) continue; const dim_t offset_r = j * num_cols; for (IType k = indptr_l[j]; k < indptr_l[j+1]; ++k) { const CType col_idx = col_idx_l[k]; -if (col_idx < seg_start || col_idx >= seg_end) continue; -const dim_t offset_out = col_idx * num_cols; -row_idx[col_idx] = 1; +if (col_idx < col_start || col_idx >= col_end) continue; + +const nnvm::dim_t rsp_row = row_flg_sum[col_idx] - 1; +const nnvm::dim_t offset_out = rsp_row * num_cols; const DType val = data_l[k]; + for (dim_t l = 0; l < num_cols; ++l) { out[offset_out+l] += data_r[offset_r+l] * val; } @@ -605,43 +608,51 @@ inline void DotCsrDnsRspImpl(const OpContext& ctx, const TBlob col_idx_l = lhs.aux_data(csr::kIdx); const TBlob& data_r = rhs; - // pre-allocate spaces for ret using the dense dimension size - ret->CheckAndAlloc({mshadow::Shape1(lhs.shape()[1])}); - const TBlob data_out = ret->data(); - const TBlob row_idx_out = ret->aux_data(rowsparse::kIdx); - MSHADOW_SGL_DBL_TYPE_SWITCH(data_l.type_flag_, DType, { // data type MSHADOW_IDX_TYPE_SWITCH(indptr_l.type_flag_, IType, { // indptr type MSHADOW_IDX_TYPE_SWITCH(col_idx_l.type_flag_, CType, { // col idx type -MSHADOW_IDX_TYPE_SWITCH(row_idx_out.type_flag_, RType, { // row idx type +MSHADOW_IDX_TYPE_SWITCH(ret->aux_type(rowsparse::kIdx), RType, { // row idx type + const dim_t num_rows = lhs.shape()[1]; + size_t workspace_size = 2 * (num_rows * sizeof(dim_t)); + mshadow::Tensor workspace = +ctx.requested[0].get_space_typed ( +mshadow::Shape1(workspace_size), s); + dim_t* row_flg = reinterpret_cast (workspace.dptr_); + dim_t* prefix_sum = row_flg + num_rows; + + Fill(s, TBlob(row_flg, mshadow::Shape1(num_rows), cpu::kDevMask), kWriteTo, 0); + mxnet_op::Kernel ::Launch(s,
[GitHub] piiswrong commented on issue #8674: ADD CapsNet example
piiswrong commented on issue #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#issuecomment-345549454 Does the results match the original paper? 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 With regards, Apache Git Services
[GitHub] piiswrong commented on a change in pull request #8674: ADD CapsNet example
piiswrong commented on a change in pull request #8674: ADD CapsNet example URL: https://github.com/apache/incubator-mxnet/pull/8674#discussion_r151877975 ## File path: example/capsnet/README.md ## @@ -0,0 +1,32 @@ +**CapsNet-MXNet** += + +This example is MXNet implementation of [CapsNet](https://arxiv.org/abs/1710.09829): +Sara Sabour, Nicholas Frosst, Geoffrey E Hinton. Dynamic Routing Between Capsules. NIPS 2017 +- The Current best test error is 0.5% + +Due to the permission issue, this example is maintained in this [repository](https://github.com/samsungsds-rnd/capsnet.mxnet) separately. Review comment: what does this mean? 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: Fixing the monitor callback of the bucketing module. (#8696)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 9dcdf57 Fixing the monitor callback of the bucketing module. (#8696) 9dcdf57 is described below commit 9dcdf575a92aa710e800f36ea070cc910ea44ee7 Author: Tobias DomhanAuthorDate: Mon Nov 20 02:32:35 2017 +0530 Fixing the monitor callback of the bucketing module. (#8696) --- python/mxnet/module/bucketing_module.py | 4 1 file changed, 4 insertions(+) diff --git a/python/mxnet/module/bucketing_module.py b/python/mxnet/module/bucketing_module.py index 4a5330e..0bea260 100644 --- a/python/mxnet/module/bucketing_module.py +++ b/python/mxnet/module/bucketing_module.py @@ -92,6 +92,7 @@ class BucketingModule(BaseModule): self._curr_module = None self._curr_bucket_key = None self._params_dirty = False +self._monitor = None def _reset_bind(self): """Internal utility function to reset binding.""" @@ -367,6 +368,8 @@ class BucketingModule(BaseModule): module.bind(data_shapes, label_shapes, self._curr_module.for_training, self._curr_module.inputs_need_grad, force_rebind=False, shared_module=self._buckets[self._default_bucket_key]) +if self._monitor is not None: +module.install_monitor(self._monitor) self._buckets[bucket_key] = module self._curr_module = self._buckets[bucket_key] @@ -510,5 +513,6 @@ class BucketingModule(BaseModule): def install_monitor(self, mon): """Installs monitor on all executors """ assert self.binded +self._monitor = mon for mod in self._buckets.values(): mod.install_monitor(mon) -- To stop receiving notification emails like this one, please contact ['"comm...@mxnet.apache.org" '].
[GitHub] piiswrong closed pull request #8696: Fixing the monitor callback of the bucketing module.
piiswrong closed pull request #8696: Fixing the monitor callback of the bucketing module. URL: https://github.com/apache/incubator-mxnet/pull/8696 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/python/mxnet/module/bucketing_module.py b/python/mxnet/module/bucketing_module.py index dd6cafb277..fa92c5d1a1 100644 --- a/python/mxnet/module/bucketing_module.py +++ b/python/mxnet/module/bucketing_module.py @@ -85,6 +85,7 @@ def __init__(self, sym_gen, default_bucket_key=None, logger=logging, self._curr_module = None self._curr_bucket_key = None self._params_dirty = False +self._monitor = None def _reset_bind(self): """Internal utility function to reset binding.""" @@ -356,6 +357,8 @@ def switch_bucket(self, bucket_key, data_shapes, label_shapes=None): module.bind(data_shapes, label_shapes, self._curr_module.for_training, self._curr_module.inputs_need_grad, force_rebind=False, shared_module=self._buckets[self._default_bucket_key]) +if self._monitor is not None: +module.install_monitor(self._monitor) self._buckets[bucket_key] = module self._curr_module = self._buckets[bucket_key] @@ -499,5 +502,6 @@ def symbol(self): def install_monitor(self, mon): """Installs monitor on all executors """ assert self.binded +self._monitor = mon for mod in self._buckets.values(): mod.install_monitor(mon) 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 With regards, Apache Git Services
[GitHub] piiswrong closed pull request #8698: [COREML] Update the json getter
piiswrong closed pull request #8698: [COREML] Update the json getter URL: https://github.com/apache/incubator-mxnet/pull/8698 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/tools/coreml/converter/_layers.py b/tools/coreml/converter/_layers.py index fe00232828..4c5ebc6fb0 100644 --- a/tools/coreml/converter/_layers.py +++ b/tools/coreml/converter/_layers.py @@ -38,6 +38,30 @@ def _get_node_name(net, node_id): def _get_node_shape(net, node_id): return net['nodes'][node_id]['shape'] +def _get_attrs(node): +"""get attribute dict from node + +This functions keeps backward compatibility +for both attr and attrs key in the json field. + +Parameters +-- +node : dict + The json graph Node + +Returns +--- +attrs : dict + The attr dict, returns empty dict if + the field do not exist. +""" +if 'attrs' in node: +return node['attrs'] +elif 'attr' in node: +return node['attr'] +else: +return {} + # TODO These operators still need to be converted (listing in order of priority): # High priority: @@ -108,7 +132,7 @@ def convert_transpose(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) axes = literal_eval(param['axes']) builder.add_permute(name, axes, input_name, output_name) @@ -180,7 +204,7 @@ def convert_activation(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -mx_non_linearity = node['attr']['act_type'] +mx_non_linearity = _get_attrs(node)['act_type'] #TODO add SCALED_TANH, SOFTPLUS, SOFTSIGN, SIGMOID_HARD, LEAKYRELU, PRELU, ELU, PARAMETRICSOFTPLUS, THRESHOLDEDRELU, LINEAR if mx_non_linearity == 'relu': non_linearity = 'RELU' @@ -281,7 +305,7 @@ def convert_convolution(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) inputs = node['inputs'] args, _ = module.get_params() @@ -361,7 +385,7 @@ def convert_pooling(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) layer_type_mx = param['pool_type'] if layer_type_mx == 'max': @@ -445,9 +469,9 @@ def convert_batchnorm(net, node, module, builder): eps = 1e-3 # Default value of eps for MXNet. use_global_stats = False # Default value of use_global_stats for MXNet. -if 'attr' in node: -if 'eps' in node['attr']: -eps = literal_eval(node['attr']['eps']) +attrs = _get_attrs(node) +if 'eps' in attrs: +eps = literal_eval(attrs['eps']) args, aux = module.get_params() gamma = args[_get_node_name(net, inputs[1][0])].asnumpy() @@ -511,7 +535,7 @@ def convert_deconvolution(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) inputs = node['inputs'] args, _ = module.get_params() 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 With regards, Apache Git Services
[incubator-mxnet] branch master updated: [COREML] Update the json getter (#8698)
This is an automated email from the ASF dual-hosted git repository. jxie pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 38a032c [COREML] Update the json getter (#8698) 38a032c is described below commit 38a032c886f56f94cdad004c89fd4e1926f85ba6 Author: Tianqi ChenAuthorDate: Sun Nov 19 12:53:19 2017 -0800 [COREML] Update the json getter (#8698) * [COREML] Update the json getter * add docstring --- tools/coreml/converter/_layers.py | 40 +++ 1 file changed, 32 insertions(+), 8 deletions(-) diff --git a/tools/coreml/converter/_layers.py b/tools/coreml/converter/_layers.py index fe00232..4c5ebc6 100644 --- a/tools/coreml/converter/_layers.py +++ b/tools/coreml/converter/_layers.py @@ -38,6 +38,30 @@ def _get_node_name(net, node_id): def _get_node_shape(net, node_id): return net['nodes'][node_id]['shape'] +def _get_attrs(node): +"""get attribute dict from node + +This functions keeps backward compatibility +for both attr and attrs key in the json field. + +Parameters +-- +node : dict + The json graph Node + +Returns +--- +attrs : dict + The attr dict, returns empty dict if + the field do not exist. +""" +if 'attrs' in node: +return node['attrs'] +elif 'attr' in node: +return node['attr'] +else: +return {} + # TODO These operators still need to be converted (listing in order of priority): # High priority: @@ -108,7 +132,7 @@ def convert_transpose(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) axes = literal_eval(param['axes']) builder.add_permute(name, axes, input_name, output_name) @@ -180,7 +204,7 @@ def convert_activation(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -mx_non_linearity = node['attr']['act_type'] +mx_non_linearity = _get_attrs(node)['act_type'] #TODO add SCALED_TANH, SOFTPLUS, SOFTSIGN, SIGMOID_HARD, LEAKYRELU, PRELU, ELU, PARAMETRICSOFTPLUS, THRESHOLDEDRELU, LINEAR if mx_non_linearity == 'relu': non_linearity = 'RELU' @@ -281,7 +305,7 @@ def convert_convolution(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) inputs = node['inputs'] args, _ = module.get_params() @@ -361,7 +385,7 @@ def convert_pooling(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) layer_type_mx = param['pool_type'] if layer_type_mx == 'max': @@ -445,9 +469,9 @@ def convert_batchnorm(net, node, module, builder): eps = 1e-3 # Default value of eps for MXNet. use_global_stats = False # Default value of use_global_stats for MXNet. -if 'attr' in node: -if 'eps' in node['attr']: -eps = literal_eval(node['attr']['eps']) +attrs = _get_attrs(node) +if 'eps' in attrs: +eps = literal_eval(attrs['eps']) args, aux = module.get_params() gamma = args[_get_node_name(net, inputs[1][0])].asnumpy() @@ -511,7 +535,7 @@ def convert_deconvolution(net, node, module, builder): """ input_name, output_name = _get_input_output_name(net, node) name = node['name'] -param = node['attr'] +param = _get_attrs(node) inputs = node['inputs'] args, _ = module.get_params() -- To stop receiving notification emails like this one, please contact ['"comm...@mxnet.apache.org" '].
[GitHub] pracheer commented on issue #8698: [COREML] Update the json getter
pracheer commented on issue #8698: [COREML] Update the json getter URL: https://github.com/apache/incubator-mxnet/pull/8698#issuecomment-345548076 Yes please! 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 With regards, Apache Git Services
[GitHub] tqchen commented on issue #8698: [COREML] Update the json getter
tqchen commented on issue #8698: [COREML] Update the json getter URL: https://github.com/apache/incubator-mxnet/pull/8698#issuecomment-345547195 Merge this ? 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 With regards, Apache Git Services
[GitHub] pracheer commented on issue #8684: [BugFix][CoreML Converter] Dense layers w/o bias.
pracheer commented on issue #8684: [BugFix][CoreML Converter] Dense layers w/o bias. URL: https://github.com/apache/incubator-mxnet/pull/8684#issuecomment-345546919 @jiajiechen @srikris 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 With regards, Apache Git Services
[GitHub] pracheer commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2?
pracheer commented on issue #8703: [DISCUSSION] (when) Should we deprecate support for python2? URL: https://github.com/apache/incubator-mxnet/issues/8703#issuecomment-345546205 I'm assuming there are still a lot of production customers who are using Python2 and will continue to do so for some foreseeable future. These customers may/will be reluctant to move to Python3 until they are really forced to (either due to lack of support for python2 or new features in python3). If we deprecate python2 too "soon", the chances are we are going to face wrath of such customers and it may not provide a happy customer experience. (Unless there is a specific feature that can't be handled w/ python2) We can let customers/developers (and rest of the python ecosystem?) move to python 3 on their own while we continue to support python2. Probably around 2019 and early 2020, we'll see a lot of people moving to python3. It may then be a good time to seriously consider deprecating python2. 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 With regards, Apache Git Services
[incubator-mxnet] 01/01: Revert "2bit gradient compression (#8662)"
This is an automated email from the ASF dual-hosted git repository. zhasheng pushed a commit to branch revert-8662-gc-pr in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git commit 2e58c0e162e081f7240db24f251d65f1d60b5f86 Author: Sheng ZhaAuthorDate: Sat Nov 18 22:59:38 2017 -0800 Revert "2bit gradient compression (#8662)" This reverts commit a499f892c9ee6f59ccfb57c9e431c91014078891. --- example/image-classification/common/fit.py | 44 ++-- example/rnn/lstm_bucketing.py | 1 + include/mxnet/c_api.h | 13 - include/mxnet/kvstore.h| 15 -- python/mxnet/gluon/trainer.py | 12 +- python/mxnet/kvstore.py| 62 - python/mxnet/module/bucketing_module.py| 17 +- python/mxnet/module/module.py | 11 +- src/c_api/c_api.cc | 14 -- src/kvstore/comm.h | 87 +-- src/kvstore/gradient_compression-inl.h | 155 src/kvstore/gradient_compression.cc| 193 -- src/kvstore/gradient_compression.cu| 40 --- src/kvstore/gradient_compression.h | 138 -- src/kvstore/kvstore.cc | 2 +- src/kvstore/kvstore_dist.h | 388 - src/kvstore/kvstore_dist_server.h | 143 ++- src/kvstore/kvstore_local.h| 7 - tests/nightly/dist_sync_kvstore.py | 120 + tests/nightly/test_kvstore.py | 200 ++- tools/bandwidth/measure.py | 6 +- 21 files changed, 167 insertions(+), 1501 deletions(-) diff --git a/example/image-classification/common/fit.py b/example/image-classification/common/fit.py index 2b002c7..51a1abe 100755 --- a/example/image-classification/common/fit.py +++ b/example/image-classification/common/fit.py @@ -103,11 +103,6 @@ def add_fit_args(parser): help='1 means test reading speed without training') train.add_argument('--dtype', type=str, default='float32', help='precision: float32 or float16') -train.add_argument('--gc-type', type=str, default='none', - help='type of gradient compression to use, \ - takes `2bit` or `none` for now') -train.add_argument('--gc-threshold', type=float, default=0.5, - help='threshold for 2bit gradient compression') return train def fit(args, network, data_loader, **kwargs): @@ -119,9 +114,6 @@ def fit(args, network, data_loader, **kwargs): """ # kvstore kv = mx.kvstore.create(args.kv_store) -if args.gc_type != 'none': -kv.set_gradient_compression({'type': args.gc_type, - 'threshold': args.gc_threshold}) # logging head = '%(asctime)-15s Node[' + str(kv.rank) + '] %(message)s' @@ -170,10 +162,10 @@ def fit(args, network, data_loader, **kwargs): lr_scheduler = lr_scheduler optimizer_params = { -'learning_rate': lr, -'wd' : args.wd, -'lr_scheduler': lr_scheduler, -'multi_precision': True} +'learning_rate': lr, +'wd' : args.wd, +'lr_scheduler': lr_scheduler, +'multi_precision': True} # Only a limited number of optimizers have 'momentum' property has_momentum = {'sgd', 'dcasgd', 'nag'} @@ -203,17 +195,17 @@ def fit(args, network, data_loader, **kwargs): # run model.fit(train, - begin_epoch= args.load_epoch if args.load_epoch else 0, - num_epoch = args.num_epochs, - eval_data = val, - eval_metric= eval_metrics, - kvstore= kv, - optimizer = args.optimizer, - optimizer_params = optimizer_params, - initializer= initializer, - arg_params = arg_params, - aux_params = aux_params, - batch_end_callback = batch_end_callbacks, - epoch_end_callback = checkpoint, - allow_missing = True, - monitor= monitor) +begin_epoch= args.load_epoch if args.load_epoch else 0, +num_epoch = args.num_epochs, +eval_data = val, +eval_metric= eval_metrics, +kvstore= kv, +optimizer = args.optimizer, +optimizer_params = optimizer_params, +initializer= initializer, +arg_params = arg_params, +aux_params = aux_params, +batch_end_callback = batch_end_callbacks, +epoch_end_callback = checkpoint, +allow_missing = True, +monitor= monitor) diff --git a/example/rnn/lstm_bucketing.py
[incubator-mxnet] branch revert-8662-gc-pr updated (eff6bb6 -> 2e58c0e)
This is an automated email from the ASF dual-hosted git repository. zhasheng pushed a change to branch revert-8662-gc-pr in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. omit eff6bb6 Revert "2bit gradient compression (#8662)" omit ff1af09 Revert "Restored some copyright attribution that were accidentally removed. (#8688)" add d2a856a Restored some copyright attribution that were accidentally removed. (#8688) add bf7a0ff Invert environment check (#8706) add 6ca944b disable test causing mklml-cpu to fail (#8713) new 2e58c0e Revert "2bit gradient compression (#8662)" This update added new revisions after undoing existing revisions. That is to say, some revisions that were in the old version of the branch are not in the new version. This situation occurs when a user --force pushes a change and generates a repository containing something like this: * -- * -- B -- O -- O -- O (eff6bb6) \ N -- N -- N refs/heads/revert-8662-gc-pr (2e58c0e) You should already have received notification emails for all of the O revisions, and so the following emails describe only the N revisions from the common base, B. Any revisions marked "omit" are not gone; other references still refer to them. Any revisions marked "discard" are gone forever. The 1 revisions listed above as "new" are entirely new to this repository and will be described in separate emails. The revisions listed as "add" were already present in the repository and have only been added to this reference. Summary of changes: amalgamation/dmlc-minimum0.cc| 1 + cmake/Modules/FindJeMalloc.cmake | 10 +- cpp-package/include/mxnet-cpp/MxNetCpp.h | 1 + cpp-package/include/mxnet-cpp/base.h | 1 + cpp-package/include/mxnet-cpp/executor.h | 1 + cpp-package/include/mxnet-cpp/initializer.h | 1 + cpp-package/include/mxnet-cpp/io.h | 1 + cpp-package/include/mxnet-cpp/kvstore.h | 1 + cpp-package/include/mxnet-cpp/lr_scheduler.h | 1 + cpp-package/include/mxnet-cpp/metric.h | 1 + cpp-package/include/mxnet-cpp/model.h| 1 + cpp-package/include/mxnet-cpp/monitor.h | 1 + cpp-package/include/mxnet-cpp/ndarray.h | 1 + cpp-package/include/mxnet-cpp/op_map.h | 1 + cpp-package/include/mxnet-cpp/op_suppl.h | 1 + cpp-package/include/mxnet-cpp/op_util.h | 1 + cpp-package/include/mxnet-cpp/operator.h | 1 + cpp-package/include/mxnet-cpp/optimizer.h| 1 + cpp-package/include/mxnet-cpp/shape.h| 1 + cpp-package/include/mxnet-cpp/symbol.h | 1 + docker/install/scala.sh | 2 +- example/image-classification/symbols/vgg.py | 16 include/mxnet/base.h | 1 + include/mxnet/c_api.h| 1 + include/mxnet/c_predict_api.h| 1 + include/mxnet/engine.h | 1 + include/mxnet/executor.h | 1 + include/mxnet/io.h | 1 + include/mxnet/kvstore.h | 1 + include/mxnet/ndarray.h | 1 + include/mxnet/op_attr_types.h| 1 + include/mxnet/operator.h | 1 + include/mxnet/operator_util.h| 1 + include/mxnet/resource.h | 1 + include/mxnet/storage.h | 1 + include/mxnet/tensor_blob.h | 1 + perl-package/AI-MXNet/lib/AI/MXNet/Types.pm | 2 +- plugin/caffe/caffe_blob.cc | 1 + plugin/caffe/caffe_blob.h| 1 + plugin/caffe/caffe_common.cc | 1 + plugin/caffe/caffe_common.h | 1 + plugin/caffe/caffe_data_iter.cc | 1 + plugin/caffe/caffe_fieldentry.h | 1 + plugin/caffe/caffe_loss-inl.h| 1 + plugin/caffe/caffe_loss.cc | 1 + plugin/caffe/caffe_loss.cu | 1 + plugin/caffe/caffe_op-inl.h | 1 + plugin/caffe/caffe_op.cc | 1 + plugin/caffe/caffe_op.cu | 1 + plugin/caffe/caffe_stream.cc | 1 + plugin/caffe/caffe_stream.h | 1 + plugin/opencv/cv_api.cc | 1 +
[incubator-mxnet] branch master updated: disable test causing mklml-cpu to fail (#8713)
This is an automated email from the ASF dual-hosted git repository. zhasheng pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git The following commit(s) were added to refs/heads/master by this push: new 6ca944b disable test causing mklml-cpu to fail (#8713) 6ca944b is described below commit 6ca944baa8b10214990a9db17cac1c34fb4a927d Author: mbaijal <30911248+mbai...@users.noreply.github.com> AuthorDate: Sun Nov 19 12:06:51 2017 -0800 disable test causing mklml-cpu to fail (#8713) --- tests/python/unittest/test_operator.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/python/unittest/test_operator.py b/tests/python/unittest/test_operator.py index 18544ba..55a3a57 100644 --- a/tests/python/unittest/test_operator.py +++ b/tests/python/unittest/test_operator.py @@ -1043,6 +1043,7 @@ def test_convolution_grouping(): np.testing.assert_allclose(arr1.asnumpy(), arr2.asnumpy(), rtol=1e-3, atol=1e-4) +@unittest.skip("test fails intermittently. temporarily disabled till it gets fixed. tracked at https://github.com/apache/incubator-mxnet/issues/8712;) def test_depthwise_convolution(): for num_base in [1, 4, 16, 32, 64]: for kernel in [(3,3), (5,5)]: -- To stop receiving notification emails like this one, please contact ['"comm...@mxnet.apache.org"'].
[GitHub] szha closed pull request #8713: Disable 'test_operator.test_depthwise_convolution' which fails in [Python2: MKLML-CPU] and [Python3: MKLML-CPU]
szha closed pull request #8713: Disable 'test_operator.test_depthwise_convolution' which fails in [Python2: MKLML-CPU] and [Python3: MKLML-CPU] URL: https://github.com/apache/incubator-mxnet/pull/8713 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/tests/python/unittest/test_operator.py b/tests/python/unittest/test_operator.py index 18544bade5..55a3a57218 100644 --- a/tests/python/unittest/test_operator.py +++ b/tests/python/unittest/test_operator.py @@ -1043,6 +1043,7 @@ def test_convolution_grouping(): np.testing.assert_allclose(arr1.asnumpy(), arr2.asnumpy(), rtol=1e-3, atol=1e-4) +@unittest.skip("test fails intermittently. temporarily disabled till it gets fixed. tracked at https://github.com/apache/incubator-mxnet/issues/8712;) def test_depthwise_convolution(): for num_base in [1, 4, 16, 32, 64]: for kernel in [(3,3), (5,5)]: 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 With regards, Apache Git Services
[GitHub] larroy commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
larroy commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345544511 @piiswrong CMake has become the de-facto standard to build C/C++ projects. Seems installing in those platforms is just make && make install. But indeed is an additional dependency, but Make is a separate program as well. Why do you say it's painful? the CMake installation doesn't seem too complex. 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 With regards, Apache Git Services
[GitHub] larroy commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
larroy commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345544304 +1 for CMake 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 With regards, Apache Git Services
[GitHub] cjolivier01 commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
cjolivier01 commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345541401 btw, CMake has the lint stuff, or more likely, a partial/outdated version of it. Targets are mxnet_lint, dmlc_lint, mshadow_lint. I also have CMake packages in another project I've made to do gcov analysis that I could pull in and maybe help the local infra ppl to add to this project. 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 With regards, Apache Git Services
[GitHub] eric-haibin-lin opened a new pull request #8717: fix group2ctx with null reqs
eric-haibin-lin opened a new pull request #8717: fix group2ctx with null reqs URL: https://github.com/apache/incubator-mxnet/pull/8717 ## Description ## It looks like the index into `arg_grad_ctxes` was not correct because `g.outputs` don't include the arg_grad whose grad_req is null. @reminisce @piiswrong @tqchen @ZiyueHuang ## Checklist ## ### Essentials ### - [x] Passed code style checking (`make lint`) - [x] Changes are complete (i.e. I finished coding on this PR) - [x] All changes have test coverage - [x] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. - [x] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [x] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here 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 With regards, Apache Git Services
[GitHub] sbodenstein commented on issue #7931: MKL-DNN integration: request for reviews
sbodenstein commented on issue #7931: MKL-DNN integration: request for reviews URL: https://github.com/apache/incubator-mxnet/pull/7931#issuecomment-345538097 @ykim362: do you know if bugs, like the resnet convergence bug, are still unsolved with v0.11 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 With regards, Apache Git Services
[GitHub] tqchen commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
tqchen commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345537323 CMake is great, and we can still use the configuration based tricks https://github.com/dmlc/tvm/blob/master/CMakeLists.txt#L10 to rely on a config.cmake to customize the build option. We might want to leave simple rules like lint to Makefile (or as a python script) since it does not really help to bring them to cmake. The only problem we might see is it increases the burden on build on embedded systems like raspberry Pi where CMake version is not great. Could be resolved by adding clear instruction to build cmake from source. 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 With regards, Apache Git Services
[GitHub] tqchen commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake?
tqchen commented on issue #8702: [DISCUSSION] Should we deprecate Makefile and only use CMake? URL: https://github.com/apache/incubator-mxnet/issues/8702#issuecomment-345537323 CMake is great, and we cans till use the configuration based tricks https://github.com/dmlc/tvm/blob/master/CMakeLists.txt#L10 to rely on a config.cmake to customize the build option. We might want to leave simple rules like lint to Makefile (or as a python script) since it does not really help to bring them to cmake. 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 With regards, Apache Git Services
[GitHub] ZiyueHuang commented on issue #8611: optimization for dot(csr.T, dense) = rsp
ZiyueHuang commented on issue #8611: optimization for dot(csr.T, dense) = rsp URL: https://github.com/apache/incubator-mxnet/pull/8611#issuecomment-345533714 Add benchmark for `n=2` case, Before, ``` python dot.py --num-omp-threads 16 mxnet sparse dot benchmark: dot(csr, default) = default (matrix multiplication: (m x k)^T * (k x n) = m x n) lhs_density(%) rhs_density(%)contextmkn t_sparse(ms) t_dense(ms) speedup 1.0 100.0 cpu(0) 256 1002 41.61 36.01 0.87 ``` After, ``` python dot.py --num-omp-threads 16 mxnet sparse dot benchmark: dot(csr, default) = default (matrix multiplication: (m x k)^T * (k x n) = m x n) lhs_density(%) rhs_density(%)contextmkn t_sparse(ms) t_dense(ms) speedup 1.0 100.0 cpu(0) 256 1002 14.44 32.46 2.25 ``` 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 With regards, Apache Git Services