Repository: incubator-systemml
Updated Branches:
refs/heads/master c1c7d3341 -> c335cd403
[SYSTEMML-1340][SYSTEMML-1341] Implemented conv2d_bias_add and relu_maxpooling
instruction for GPU
Closes #425.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 609649484 -> 8c1e89e51
[HOTFIX] Show total time in statistics for MLContext
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 95f300d9d -> 4344ad0c6
[MINOR] Added common errors and troubleshooting tricks
Closes #428.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/gh-pages 42e86e76c -> bd232241b
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/alg-ref/SystemML_Algorithms_Reference.tex
--
diff --git
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/alg-ref/DescriptiveBivarStats.tex
--
diff --git a/alg-ref/DescriptiveBivarStats.tex
b/alg-ref/DescriptiveBivarStats.tex
new file mode 100644
index
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/Algorithms
Reference/DescriptiveBivarStats.tex
--
diff --git a/Algorithms Reference/DescriptiveBivarStats.tex b/Algorithms
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/alg-ref/GLM.tex
--
diff --git a/alg-ref/GLM.tex b/alg-ref/GLM.tex
new file mode 100644
index 000..8555a5b
--- /dev/null
+++ b/alg-ref/GLM.tex
@@ -0,0
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/Algorithms
Reference/GLM.tex
--
diff --git a/Algorithms Reference/GLM.tex b/Algorithms Reference/GLM.tex
deleted file mode 100644
index 8555a5b..000
---
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/Algorithms
Reference/SystemML_Algorithms_Reference.tex
--
diff --git a/Algorithms Reference/SystemML_Algorithms_Reference.tex
b/Algorithms
[MINOR] Added common errors and troubleshooting tricks
Closes #428.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/bd232241
Tree:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/Algorithms
Reference/LinReg.tex
--
diff --git a/Algorithms Reference/LinReg.tex b/Algorithms Reference/LinReg.tex
deleted file mode 100644
index
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/358cfc9f/alg-ref/LinReg.tex
--
diff --git a/alg-ref/LinReg.tex b/alg-ref/LinReg.tex
new file mode 100644
index 000..67273c2
--- /dev/null
+++ b/alg-ref/LinReg.tex
Repository: incubator-systemml
Updated Branches:
refs/heads/master 20e33bbe8 -> eef34599e
[HOTFIX] Bugfix in setGPU method of MLContext
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/eef34599
Repository: systemml
Updated Branches:
refs/heads/master ca5581fcc -> b2700839b
[SYSTEMML-540] [MINOR] Bugfix for SoftMaxWithLoss layer in Caffe2DML
- This commit allows users to specify the bottom of SoftMaxWithLoss in
any order they prefer, thereby avoiding subtle bugs
Project:
Repository: systemml
Updated Branches:
refs/heads/master dc4bfd95e -> 81b9248fc
[MINOR] Resolve race condition between locking of metastore_db of Scala
SparkSession and PySpark SparkSession when using SystemML MLContext API
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
Repository: systemml
Updated Branches:
refs/heads/master 8707ff1d4 -> f046051d4
[HOTFIX] [SYSTEMML-540] Bugfix in validation of convolution operations.
- Also, updated the DML documentation as per Prithvi's recommendation.
- This hotfix also contains minor bugfix in getP, getQ methods of
[SYSTEMML-1796] Provide fine-grained statistics for SystemML
Though SystemML provides explain and statistics tooling functionality to
understand the execution plan and the performance, most users (even
committers) are not able to explain why SystemML's performance is so good
or bad. This is
Repository: systemml
Updated Branches:
refs/heads/master 6133be23e -> 648eb21d6
http://git-wip-us.apache.org/repos/asf/systemml/blob/648eb21d/src/main/java/org/apache/sysml/runtime/matrix/data/MatrixBlock.java
--
diff --git
http://git-wip-us.apache.org/repos/asf/systemml/blob/648eb21d/src/main/java/org/apache/sysml/runtime/instructions/cp/ConvolutionCPInstruction.java
--
diff --git
Repository: systemml
Updated Branches:
refs/heads/master eee35e984 -> 1f5b14dda
[SYSTEMML-540] Allow mllearn models to load the model eagerly.
- This simplifies performance debugging of training and scoring.
Closes #574.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
Repository: systemml
Updated Branches:
refs/heads/master 1a3d85f91 -> 8f412ac5c
[SYSTEMML-1798] Make Python MLContext API and Scala/Java MLContext API
consistent in terms of functionality and naming
- Provide getScriptExecutionString and getScriptString for Python Script object.
- The Python
Repository: systemml
Updated Branches:
refs/heads/master ee2b37e4f -> 11b689d49
[SYSTEMML-1802] Fix build failure for distribution profile on Windows
Closes #594.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
[SYSTEMML-1798] Make Python MLContext API and Scala/Java MLContext API
consistent in terms of functionality and naming
- Provide getScriptExecutionString and getScriptString for Python Script object.
- The Python API has no corresponding objects for ScriptExecutor,
MatrixMetadata and
[SYSTEMML-1768] Cleanup properties of systemml-config file
This patch cleans up the following two properties of the
SystemML-config.xml file in order to better convey their meaning:
1) cp.parallel.matrixmult -> cp.parallel.ops
2) cp.parallel.textio -> cp.parallel.io
Project:
[SYSTEMML-540] Extended Caffe2DML to support image segmentation problems
- This commit extends Caffe2DML to support image segmentation problem,
depthwise convolution and has couple of bugfixes regarding loading
existing caffe model.
- Additionally, we have added a summary() method to
Repository: systemml
Updated Branches:
refs/heads/gh-pages 05792e0e9 -> 0ff267404
[SYSTEMML-540] Extended Caffe2DML to support image segmentation problems
- This commit extends Caffe2DML to support image segmentation problem,
depthwise convolution and has couple of bugfixes regarding
Repository: systemml
Updated Branches:
refs/heads/master d56c05ece -> 978d4de47
http://git-wip-us.apache.org/repos/asf/systemml/blob/978d4de4/src/main/java/org/apache/sysml/parser/common/CommonSyntacticValidator.java
--
diff
Repository: incubator-systemml
Updated Branches:
refs/heads/master 2a4e4f8b3 -> 39351749f
[HOTFIX] Bugfix for bias_add and bias_multiply operations
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 32924dc60 -> 9ed27ad60
[SYSTEMML-1552] Support GPU via Python MLContext API
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 32f075695 -> 2d2196d84
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/2d2196d8/src/test/scripts/functions/tensor/Conv2DTest.dml
--
diff --git
[SYSTEMML-687] Optimized LibMatrixDNN for sparse inputs
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/2d2196d8
Tree: http://git-wip-us.apache.org/repos/asf/incubator-systemml/tree/2d2196d8
Diff:
Repository: systemml
Updated Branches:
refs/heads/master 257bf3ed0 -> 9da5eab00
[MINOR] Bugfix to resolve UnboundLocalError in mlcontext
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit: http://git-wip-us.apache.org/repos/asf/systemml/commit/9da5eab0
Tree:
Repository: systemml
Updated Branches:
refs/heads/master 0325da7de -> 4d5a82ecf
[SYSTEMML-540] Avoid unnecessary sparse-to-dense conversion on depthwise
convolution layers and probability matrix
```
Network: unet
Setup: 30g driver, K80 GPU (only 1 GPU used)
Performance before this commit:
Repository: systemml
Updated Branches:
refs/heads/master 4d5a82ecf -> 3ca053535
[SYSTEMML-445] Integrate GPU exectype selection into our existing infrastructure
Closes #627.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
usercontent.com/niketanpansare/model_zoo/master/caffe/vision/lenet/mnist/lenet.proto',
'lenet.proto')
+urllib.urlretrieve('https://raw.githubusercontent.com/niketanpansare/model_zoo/master/caffe/vision/lenet/mnist/lenet_solver.proto',
'lenet_solver.proto')
+
+# Train Lenet On MNIST using scikit-learn
Split the data into training and test
+n_samples = len(X)
+X_train = X[:int(.9 * n_samples)]
+y_train = y[:int(.9 * n_samples)]
+X_test = X[int(.9 * n_samples):]
+y_test = y[int(.9 * n_samples):]
+
+# Download the Lenet network
+import urllib
+urllib.urlretrieve('https://raw.githubusercontent.com/niketanpa
Repository: incubator-systemml
Updated Branches:
refs/heads/gh-pages c5ff65305 -> 1efaa5520
[SYSTEMML-769] Adding support for native BLAS for Linux
- Both MKL and OpenBLAS show 2-3x performance benefits on conv2d
operators on Lenet.
- There are several OpenMP related issues on Mac, hence we
Repository: incubator-systemml
Updated Branches:
refs/heads/master 7989ab4f3 -> 8324b69f1
[HOTFIX] Allows multiple MLContext to set the configuration property
- Also, added bugfix in mllearn to enable force GPU option.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Repository: incubator-systemml
Updated Branches:
refs/heads/master 9ee569e0f -> 59ff8b3d4
[HOTFIX] [SYSTEMML-540] Enabled commandline statistics and disabled native by
default
- Fixed a bug that disabled statistics when SystemML is invoked from
commandline.
- When change was made from
Repository: incubator-systemml
Updated Branches:
refs/heads/master 59ff8b3d4 -> 7ed8e3f49
[SYSTEMML-1479] Integrated DeConvolution and Concat layer in Caffe2DML
Closes #492.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 7ba17c7f6 -> fd41f0253
[SYSTEMML-1630] Remove confusing warning for disabled native BLAS
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
erriding ufunc, but this
should be enabled in next release.
-Until then to test above code, please use:
-
-```bash
-git clone https://github.com/niketanpansare/numpy.git
-cd numpy
-python setup.py install
-```
+This class also supports several input/output formats such as NumPy arrays,
Pandas
Repository: incubator-systemml
Updated Branches:
refs/heads/gh-pages 85137b178 -> 1246604b0
[MINOR] Fixed the format in the native-backend documentation
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
gt;> from pyspark.sql import SQLContext
->>> sqlCtx = SQLContext(sc)
>>> from mlxtend.data import mnist_data
>>> import numpy as np
>>> from sklearn.utils import shuffle
@@ -603,25 +689,23 @@ class Caffe2DML(BaseSystemMLClassifier):
>&
: 500`: Display validation loss after every 500 iterations.
+- `test_iter: 10`: Validation data size = 10 * BATCH_SIZE.
+
+
+```python
from mlxtend.data import mnist_data
import numpy as np
from sklearn.utils import shuffle
+import urllib
+from systemml.mllearn import Caffe2DML
+
+# Download t
mma ^ iter`).
+- `display: 100`: Display training loss after every 100 iterations.
+- `test_interval: 500`: Display validation loss after every 500 iterations.
+- `test_iter: 10`: Validation data size = 10 * BATCH_SIZE.
+
+
+```python
from mlxtend.data import mnist_data
import numpy as np
from skl
Repository: incubator-systemml
Updated Branches:
refs/heads/gh-pages 8b4c2b7a8 -> c067a5858
[SYSTEMML-1629] Code format for IntelliJ
Add code style file for IntelliJ so that annotation formatting can be
handled in a similar fashion for both Eclipse and IntelliJ.
Closes #515.
Project:
riding ufunc, but this
should be enabled in next release.
-Until then to test above code, please use:
+The matrix class also supports most of NumPy's universal functions (i.e.
ufuncs):
```bash
-git clone https://github.com/niketanpansare/numpy.git
-cd numpy
-python setup.py install
+pip inst
Repository: incubator-systemml
Updated Branches:
refs/heads/master 28c92b93f -> 19eed8f38
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/19eed8f3/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixDNNHelper.java
Repository: incubator-systemml
Updated Branches:
refs/heads/master 7ed36a98f -> 88f4a468f
[SYSTEMML-1661] Added the documentation for bias_add and bias_multiply
builtin function.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
[SYSTEMML-1661] Added the documentation for bias_add and bias_multiply
builtin function.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/cc0988bf
Tree:
Repository: systemml
Updated Branches:
refs/heads/gh-pages 2cfd77491 -> 7c4907095
[SYSTEMML-540] Allow for `allreduce_parallel_batches` for multi-GPU training
- Introduced `allreduce_parallel_batches` for multi-GPU training as per
Mike's suggestion.
- Moved `train_algo` and `test_algo` from
Repository: systemml
Updated Branches:
refs/heads/master ddcb9e019 -> 933a17c5f
[SYSTEMML-540] Allow for `allreduce_parallel_batches` for multi-GPU training
- Introduced `allreduce_parallel_batches` for multi-GPU training as per
Mike's suggestion.
- Moved `train_algo` and `test_algo` from
Repository: incubator-systemml
Updated Branches:
refs/heads/master 576eb4e96 -> 686363208
[SYSTEMML-1589] Check for the empty block case
Closes #487.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 19e21744c -> 6c215e700
[SYSTEMML-1573] Incorporate ALLOW_OPERATOR_FUSION in ConvolutionOp for
developer testing
Closes #482.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 738ddfc17 -> b50fe1078
[SYSTEMML-540] Added the documentation for convolution
Closes #464.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master d0ab2af19 -> 7a3e95716
[SYSTEMML-591] Initial attempt to consolidate replicated compilation chain
This only consolidates MLContext and DMLScript, but not JMLC and old
MLContext (as it is going to be removed soon).
Closes
Repository: incubator-systemml
Updated Branches:
refs/heads/master 39a37ae40 -> 738ddfc17
[SYSTEMML-692] Separated Caffe2dml into a separate jar
1. Caffe2DML should be integrated in main SystemML jar. Pros:
One jar for all. Cons: +1 MB.
2. Caffe2DML should be released as a separate jar. Pros:
[SYSTEMML-1553] Add Caffe and Tensorflow Licenses based on discussion in the
mailing list
Closes #467.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/d0ab2af1
Tree:
Repository: incubator-systemml
Updated Branches:
refs/heads/master b50fe1078 -> d0ab2af19
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/d0ab2af1/src/main/python/LICENSE
--
diff --git a/src/main/python/LICENSE
Repository: incubator-systemml
Updated Branches:
refs/heads/master abc9686fb -> b9814ccf0
[SYSTEMML-540] Additional tests to compare the accuracy of different
convolution related operators with CuDNN
Closes #477.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
[SYSTEMML-769] Adding support for native BLAS for Linux
- Both MKL and OpenBLAS show 2-3x performance benefits on conv2d
operators on Lenet.
- There are several OpenMP related issues on Mac, hence we have explicitly
disabled native support for Mac and Windows. Since we have already have
Repository: incubator-systemml
Updated Branches:
refs/heads/master 4b6d468bb -> 39a37ae40
http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/39a37ae4/src/test/java/org/apache/sysml/test/integration/AutomatedTestBase.java
Repository: incubator-systemml
Updated Branches:
refs/heads/master 4e445d17e -> 29c307c9a
[SYSTEMML-1580] Remove the usage of SYSTEMML_BLAS environment variable from
NativeHelper
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 29c307c9a -> 76f3ca5d3
[HOTFIX] Bugfix for metadata of conv2d_* and maxpool_* operations
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
Repository: incubator-systemml
Updated Branches:
refs/heads/master 259742c40 -> 700b08094
[MINOR] Added documentation for Caffe2DML APIs.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo
Commit:
http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/700b0809
Repository: systemml
Updated Branches:
refs/heads/master 3acd94186 -> e624d149f
[SYSTEMML-540] Improve the performance of CP Convolution operators
- Support sparse bias_multiply.
- Allow JIT to optimize loops in maxpooling operations.
- Perform examSparsity to optimize the future
http://git-wip-us.apache.org/repos/asf/systemml/blob/f07b5a2d/src/main/scala/org/apache/sysml/api/dl/CaffeLayer.scala
--
diff --git a/src/main/scala/org/apache/sysml/api/dl/CaffeLayer.scala
Repository: systemml
Updated Branches:
refs/heads/master ebb6ea612 -> f07b5a2d9
http://git-wip-us.apache.org/repos/asf/systemml/blob/f07b5a2d/src/main/scala/org/apache/sysml/api/dl/Utils.scala
--
diff --git
http://git-wip-us.apache.org/repos/asf/systemml/blob/f07b5a2d/src/main/scala/org/apache/sysml/api/dl/CaffeNetwork.scala
--
diff --git a/src/main/scala/org/apache/sysml/api/dl/CaffeNetwork.scala
[SYSTEMML-540] Support loading of batch normalization weights in .caffemodel
file using Caffe2DML
- Also fixed scala formatting.
Closes #662.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit: http://git-wip-us.apache.org/repos/asf/systemml/commit/f07b5a2d
Tree:
[SYSTEMML-445] Improved performance of GPU right indexing
- Improved performance of dense right-indexing by increasing parallelization to
elements rather than row.
- Added slice_sparse_dense_nnz to handle indexing of sparse wide matrices.
- Added EAGER_CUDA_FREE flag to allow users to
http://git-wip-us.apache.org/repos/asf/systemml/blob/34bb3ca8/src/main/cpp/kernels/SystemML.ptx
--
diff --git a/src/main/cpp/kernels/SystemML.ptx
b/src/main/cpp/kernels/SystemML.ptx
index 54b53b9..b4a6559 100644
---
Repository: systemml
Updated Branches:
refs/heads/master c14682b9c -> 34bb3ca82
http://git-wip-us.apache.org/repos/asf/systemml/blob/34bb3ca8/src/main/java/org/apache/sysml/api/DMLScript.java
--
diff --git
Repository: systemml
Updated Branches:
refs/heads/master 0a984a43b -> c6d499d3e
[MINOR] Refactored the locks to seperate out read and write lock
- Refactoring the locks will avoid future bugs where the developer tries
to obtain 2 write lock or a read lock on a write-locked objects, etc.
- I
[SYSTEMML-445] Improved performance of GPU right indexing
- Added slice_dense_dense kernel for right indexing
- Fixed GPU bufferpool performance bug where we were doing unnecessary
recomputation of nnz for sparse input.
- Refactored asserts to throw exceptions to be consistent with our other
http://git-wip-us.apache.org/repos/asf/systemml/blob/4cf95c92/src/main/cpp/kernels/SystemML.ptx
--
diff --git a/src/main/cpp/kernels/SystemML.ptx
b/src/main/cpp/kernels/SystemML.ptx
index caa3fc7..54b53b9 100644
---
Repository: systemml
Updated Branches:
refs/heads/master e624d149f -> 4cf95c92e
http://git-wip-us.apache.org/repos/asf/systemml/blob/4cf95c92/src/main/java/org/apache/sysml/api/DMLScript.java
--
diff --git
[SYSTEMML-1923] Support sinh, cosh and tanh as built-in functions
- Added sinh, cosh and tanh builtin functions to CP, Spark and GPU backend.
- Added these functions to Python DSL.
- Also, updated the DML language reference and Python reference documentation.
Closes #668.
Project:
[SYSTEMML-1925] Fix consistency SystemML configuration properties
This patch creates consistency in terms of a common prefix for all
configuration properties. Before we used the prefixes dml, systemml, or
none, which is now consolidated to sysml.
Project:
Repository: systemml
Updated Branches:
refs/heads/master aa15197ec -> 50a895f86
http://git-wip-us.apache.org/repos/asf/systemml/blob/50a895f8/src/main/java/org/apache/sysml/hops/codegen/cplan/CNodeUnary.java
--
diff --git
[SYSTEMML-1923] Support sinh, cosh and tanh as built-in functions
- Added sinh, cosh and tanh builtin functions to CP, Spark and GPU backend.
- Added these functions to Python DSL.
- Also, updated the DML language reference and Python reference documentation.
Closes #668.
Project:
Repository: systemml
Updated Branches:
refs/heads/master 1f3dfbc19 -> 374b0b197
[SYSTEMML-1793] [MINOR] GPU right indexing bugfix
- Since GPU only supports sparse_dense rightindexing, updated the memory
budget accordingly.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
Repository: systemml
Updated Branches:
refs/heads/master 374b0b197 -> 355373990
[SYSTEMML-1940] Modify the default directory of performance test to avoid maven
compilation issue
Closes #675.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
Repository: systemml
Updated Branches:
refs/heads/master e106966a9 -> c69bd7f30
[SYSTEMML-446] Allow SystemML to run in -exec singlenode with gpu
Currently, we can either force all instructions to:
- CP (via -exec singlenode) OR
- GPU (via -gpu force)
However, a typical user may want to run
Repository: systemml
Updated Branches:
refs/heads/master 2ca2d8aa7 -> 5adb330de
[SYSTEMML-540] Reduce the number of unknowns in ConvolutionOp
- This commit reduces the unknowns during dynamic recompilation by inferring the
input's height/width of ConvolutionOp based on its parent's output's
Repository: systemml
Updated Branches:
refs/heads/master 8395ffb29 -> 3b8a86065
[SYSTEMML-540] [MINOR] Throw user-friendly message for invalid convolution
parameters
- This commit checks if the given parameters generates an output
activation of negative dimensions.
Project:
Repository: systemml
Updated Branches:
refs/heads/master bd1946a3d -> 4744da167
[SYSTEMML-540] Support sparse output for CP im2col operator
- We support sparse output for im2col, which improved the performance by
4x on CPU for 1-layer CNN with input sparsity of 0.04.
- This commit also
Repository: systemml
Updated Branches:
refs/heads/master 0ba9e74b9 -> b34079a28
[MINOR] Enable systemml to be imported in the pyspark workers
- Moved the race condition avoidance logic to classloader instead of
import level. This avoids creation of dataframe in pyspark workers.
Closes
http://git-wip-us.apache.org/repos/asf/systemml/blob/772d9302/src/main/cpp/kernels/SystemML.ptx
--
diff --git a/src/main/cpp/kernels/SystemML.ptx
b/src/main/cpp/kernels/SystemML.ptx
index f6ba15a..caa3fc7 100644
---
Repository: systemml
Updated Branches:
refs/heads/master a0cf8e3be -> 772d9302d
http://git-wip-us.apache.org/repos/asf/systemml/blob/772d9302/src/main/java/org/apache/sysml/runtime/matrix/data/LibMatrixCuDNN.java
--
diff --git
http://git-wip-us.apache.org/repos/asf/systemml/blob/772d9302/src/main/java/org/apache/sysml/hops/ConvolutionOp.java
--
diff --git a/src/main/java/org/apache/sysml/hops/ConvolutionOp.java
[SYSTEMML-540] Support sparse GPU conv2d as well as fix memory estimation of
convolution operations
Design doc: Memory estimation of GPU operators
- Since not all operator are supported on GPU, isGPUEnabled indicates
whether an operation is enabled for GPU. This method doesnot take into
account
Repository: systemml
Updated Branches:
refs/heads/master 722c8f155 -> 5a2b49f20
[SYSTEMML-1891] Prevent 'Script string is blank' exception for Python
Closes #659.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
[SYSTEMML-1493] [SYSTEMML-1500] Added TanH and Euclidean Loss in Caffe2DML
- Added the reference documentation
- Added a new layer softmax_loss.dml
- Added compute_loss_accuracy to l2_loss.dml
- Updated tanh layer to invoke newly added builtin function
Closes #672.
Project:
[SYSTEMML-1493] [SYSTEMML-1500] Added TanH and Euclidean Loss in Caffe2DML
- Added the reference documentation
- Added a new layer softmax_loss.dml
- Added compute_loss_accuracy to l2_loss.dml
- Updated tanh layer to invoke newly added builtin function
Closes #672.
Project:
Repository: systemml
Updated Branches:
refs/heads/master b5ef21fdc -> 61dcc85e4
http://git-wip-us.apache.org/repos/asf/systemml/blob/61dcc85e/src/main/scala/org/apache/sysml/api/dl/CaffeLayer.scala
--
diff --git
Repository: systemml
Updated Branches:
refs/heads/gh-pages e3f0cf404 -> 91190d6df
[SYSTEMML-1928] Removed Tensorboard visualization code from Caffe2DML
- This avoids compilation error on powerpc.
Closes #671.
Project: http://git-wip-us.apache.org/repos/asf/systemml/repo
Commit:
Repository: systemml
Updated Branches:
refs/heads/master 61dcc85e4 -> 0cb2f7f68
[MINOR] [SYSTEMML-446] Added time spent in jcuda sync to fine-grained statistics
- Also added force accelerator flag to LibMatrixCuDNN to skip worst-case memory
budget restriction.
Project:
Repository: systemml
Updated Branches:
refs/heads/master e4c74eda6 -> 43b573dfb
[SYSTEMML-445] Select CuDNN operator dynamically based on memory budget
- Refactored CuDNN operator selection logic into separate file to simplify
LibMatrixCuDNN.
- Fixed a minor memory leak in conv2d backward
101 - 200 of 395 matches
Mail list logo