Repository: incubator-singa Updated Branches: refs/heads/master 3a64342d0 -> 8cf18e5b0
update the docs of schedule for v1.1 and installation for a FAQ entry Project: http://git-wip-us.apache.org/repos/asf/incubator-singa/repo Commit: http://git-wip-us.apache.org/repos/asf/incubator-singa/commit/8cf18e5b Tree: http://git-wip-us.apache.org/repos/asf/incubator-singa/tree/8cf18e5b Diff: http://git-wip-us.apache.org/repos/asf/incubator-singa/diff/8cf18e5b Branch: refs/heads/master Commit: 8cf18e5b069b1093026a583a0011af20225dc7ca Parents: 3a64342 Author: Wei Wang <[email protected]> Authored: Thu Oct 6 15:13:07 2016 +0800 Committer: Wei Wang <[email protected]> Committed: Thu Oct 6 15:14:36 2016 +0800 ---------------------------------------------------------------------- doc/en/develop/schedule.rst | 74 ++++++++++++++++++++++------------------ doc/en/docs/installation.md | 29 ++++++++++++++-- 2 files changed, 67 insertions(+), 36 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/incubator-singa/blob/8cf18e5b/doc/en/develop/schedule.rst ---------------------------------------------------------------------- diff --git a/doc/en/develop/schedule.rst b/doc/en/develop/schedule.rst index ef51496..c097407 100644 --- a/doc/en/develop/schedule.rst +++ b/doc/en/develop/schedule.rst @@ -20,38 +20,44 @@ Development Schedule ==================== .. csv-table:: - :header: "Release", "Module", "Feature", "Status" + :header: "Release","Module","Feature" - " 0.1 Sep 2015 "," Neural Network "," Feed forward neural network, including CNN, MLP "," done " - " "," "," RBM-like model, including RBM "," done " - " "," "," Recurrent neural network, including standard RNN "," done " - " "," Architecture "," One worker group on single node (with data partition) "," done " - " "," "," Multi worker groups on single node using [Hogwild](http://www.eecs.berkeley.edu/~brecht/papers/hogwildTR.pdf) ","done" - " "," "," Distributed Hogwild","done" - " "," "," Multi groups across nodes, like [Downpour](http://papers.nips.cc/paper/4687-large-scale-distributed-deep-networks) ","done" - " "," "," All-Reduce training architecture like [DeepImage](http://arxiv.org/abs/1501.02876) ","done" - " "," "," Load-balance among servers "," done" - " "," Failure recovery "," Checkpoint and restore ","done" - " "," Tools "," Installation with GNU auto tools"," done" - "0.2 Jan 2016 "," Neural Network "," Feed forward neural network, including AlexNet, cuDNN layers, etc."," done " - " "," "," Recurrent neural network, including GRULayer and BPTT","done " - " "," "," Model partition and hybrid partition","done" - " "," Tools "," Integration with Mesos for resource management","done" - " "," "," Prepare Docker images for deployment","done" - " "," "," Visualization of neural net and debug information ","done" - " "," Binding "," Python binding for major components ","done" - " "," GPU "," Single node with multiple GPUs ","done" - "0.3 April 2016 "," GPU "," Multiple nodes, each with multiple GPUs","done" - " "," "," Heterogeneous training using both GPU and CPU [CcT](http://arxiv.org/abs/1504.04343)","done" - " "," "," Support cuDNN v4 "," done" - " "," Installation "," Remove dependency on ZeroMQ, CZMQ, Zookeeper for single node training","done" - " "," Updater "," Add new SGD updaters including Adam, AdamMax and AdaDelta","done" - " "," Binding "," Enhance Python binding for training","done" - "1.0 Sep 2016 "," Programming abstraction ","Tensor with linear algebra, neural net and random operations "," " - " "," ","Updater for distributed parameter updating ","" - " "," Hardware "," Use Cuda and Cudnn for Nvidia GPU","" - " "," "," Use OpenCL for AMD GPU or other devices","" - " "," Cross-platform "," To extend from Linux to MacOS","" - " "," Examples "," Speech recognition example","" - " "," ","Large image models, e.g., [VGG](https://arxiv.org/pdf/1409.1556.pdf) and [Residual Net](http://arxiv.org/abs/1512.03385)","" - "1.1 Dec 2016 "," ",""," " + "0.1 Sep 2015 ","Neural Network ","Feed forward neural network, including CNN, MLP " + " "," ","RBM-like model, including RBM " + " "," ","Recurrent neural network, including standard RNN " + " ","Architecture ","One worker group on single node (with data partition) " + " "," ","Multi worker groups on single node using `Hogwild <http://www.eecs.berkeley.edu/~brecht/papers/hogwildTR.pdf>`_ " + " "," ","Distributed Hogwild" + " "," ","Multi groups across nodes, like `Downpour <http://papers.nips.cc/paper/4687-large-scale-distributed-deep-networks>`_" + " "," ","All-Reduce training architecture like `DeepImage <http://arxiv.org/abs/1501.02876>`_ " + " "," ","Load-balance among servers " + " ","Failure recovery ","Checkpoint and restore " + " ","Tools ","Installation with GNU auto Tools " + "0.2 Jan 2016 ","Neural Network ","Feed forward neural network, including AlexNet, cuDNN layers,Tools " + " "," ","Recurrent neural network, including GRULayer and BPTT " + " "," ","Model partition and hybrid partition " + " ","Tools ","Integration with Mesos for resource management " + " "," ","Prepare Docker images for deployment" + " "," ","Visualization of neural net and debug information " + " ","Binding ","Python binding for major components " + " ","GPU ","Single node with multiple GPUs " + "0.3 April 2016 ","GPU ","Multiple nodes, each with multiple GPUs" + " "," ","Heterogeneous training using both GPU and CPU `CcT <http://arxiv.org/abs/1504.04343>`_" + " "," ","Support cuDNN v4 " + " ","Installation ","Remove dependency on ZeroMQ, CZMQ, Zookeeper for single node training" + " ","Updater ","Add new SGD updaters including Adam, AdamMax and AdaDelta" + " ","Binding ","Enhance Python binding for training" + "1.0 Sep 2016 ","Programming abstraction ","Tensor with linear algebra, neural net and random operations " + " "," ","Updater for distributed parameter updating " + " ","Hardware ","Use Cuda and Cudnn for Nvidia GPU" + " "," ","Use OpenCL for AMD GPU or other devices" + " ","Cross-platform ","To extend from Linux to MacOS" + " "," ","Large image models, e.g., `VGG <https://arxiv.org/pdf/1409.1556.pdf>`_ and `Residual Net <http://arxiv.org/abs/1512.03385>`_" + "1.1 Dec 2016 ","Model Zoo ","Health-care models and popular image models" + " ","Caffe converter ","Use SINGA to train models configured in caffe proto files" + " ","Memory optimization ","Replace CNMEM with new memory pool to reduce memory footprint" + " ","Distributed training ","Migrate distributed training frameworks from V0.3" + " ","Compilation and installation ","Windows suppport" + " "," ","Simplify the installation by compiling protobuf and openblas together with SINGA" + " "," ","Build python wheel automatically using Jenkins" + " "," ","Deploy SINGA programs on Android phones for prediction tasks" http://git-wip-us.apache.org/repos/asf/incubator-singa/blob/8cf18e5b/doc/en/docs/installation.md ---------------------------------------------------------------------- diff --git a/doc/en/docs/installation.md b/doc/en/docs/installation.md index e18447b..e4b76c0 100755 --- a/doc/en/docs/installation.md +++ b/doc/en/docs/installation.md @@ -37,7 +37,7 @@ The following instructions are tested on Ubuntu 14.04 for installing dependent l # optional libraries $ sudo apt-get install python2.7-dev python-pip python-numpy - $ sudo apt-get install llibopencv-dev ibgoogle-glog-dev liblmdb-dev + $ sudo apt-get install libopencv-dev libgoogle-glog-dev liblmdb-dev Please note that PySINGA requires swig >=3.0, which could be installed via apt-get on Ubuntu 16.04; but it has to be installed from source for other Ubuntu versions including 14.04. @@ -68,6 +68,7 @@ To let the runtime know the openblas path, please export ### pip and anaconda for PySINGA pip and anaconda could be used to install python packages, e.g. numpy. +Python virtual environment is recommended to run PySINGA. To use pip with virtual environment, # install virtualenv @@ -219,6 +220,30 @@ To be added. ## FAQ +* Q: Error from 'import singa' using PySINGA installed from wheel. + + A: Please check the detailed error from `python -c "from singa import _singa_wrap"`. Sometimes it is + caused by the dependent libraries, e.g. there are multiple versions of protobuf or missing of cudnn. Following + steps show the solutions for different cases + 1. check the cudnn and cuda and gcc versions, cudnn5 and cuda7.5 and gcc4.8/4.9 are preferred. if gcc is 5.0, then downgrade it. + if cudnn is missing or not match with the wheel version, you can download the correct version of cudnn into ~/local/cudnn/ and + ``` + echo "export LD_LIBRARY_PATH=/home/<yourname>/local/cudnn/lib64:$LD_LIBRARY_PATH" >> ~/.bashrc + ``` + 2. if it is the problem related to protobuf, then better install protobuf from source into a local folder, say ~/local/; + Decompress the tar file, and then + ``` + ./configure --prefix=/home/<yourname>local + make && make install + echo "export LD_LIBRARY_PATH=/home/<yourname>/local/lib:$LD_LIBRARY_PATH" >> ~/.bashrc + source ~/.bashrc + 3. if it cannot find other libs including python, then please create virtual env using pip or conda; + and then install SINGA via + ``` + pip install --upgrade <url of singa wheel> + ``` + + * Q: Error from running `cmake ..`, which cannot find the dependent libraries. A: If you haven't installed the libraries, please install them. If you installed @@ -276,7 +301,7 @@ To be added. * Q: When I build protocol buffer, it reports that GLIBC++_3.4.20 not found in /usr/lib64/libstdc++.so.6. - A9: This means the linker found libstdc++.so.6 but that library + A: This means the linker found libstdc++.so.6 but that library belongs to an older version of GCC than was used to compile and link the program. The program depends on code defined in the newer libstdc++ that belongs to the newer version of GCC, so the linker
