Ag fix9060 (#519)

* Development updates (deeplearning4j#9053) * RL4J: Add generic update rule (#502) Signed-off-by: Alexandre Boulanger <aboulang2002@yahoo.com> * Shyrma reduce (#481) * - start working on improving of cpu legacy code for reduce ops Signed-off-by: Yurii <iuriish@yahoo.com> * - further work on improving legacy loops Signed-off-by: Yurii <iuriish@yahoo.com> * - still working on improving reduce ops Signed-off-by: Yurii <iuriish@yahoo.com> * - further work on improving reduce ops Signed-off-by: Yurii <iuriish@yahoo.com> * - testing speed run of new reduce op Signed-off-by: Yurii <iuriish@yahoo.com> * - working on improvement of default loop for reduce op Signed-off-by: Yurii <iuriish@yahoo.com> * - update signatures of stuff which calls reduce ops Signed-off-by: Yurii <iuriish@yahoo.com> * - make corrections in cuda reduce kernels Signed-off-by: Yurii <iuriish@yahoo.com> * - change loop for default case in broadcast legacy ops Signed-off-by: Yurii <iuriish@yahoo.com> * - comment some shape stuff Signed-off-by: Yurii <iuriish@yahoo.com> * - comment unnecessary prints in RNGtests Signed-off-by: Yurii <iuriish@yahoo.com> * - finish to resolve conflicts after master has been merged Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of some compilation mistakes of cuda stuff Signed-off-by: Yurii <iuriish@yahoo.com> * - minor changes Signed-off-by: Yurii <iuriish@yahoo.com> * - further search for bug causing crash on java test Signed-off-by: Yurii <iuriish@yahoo.com> * - add scalar case in reduce_ ... exec stuff Signed-off-by: Yurii <iuriish@yahoo.com> * - minor corrections in NAtiveOps.cu Signed-off-by: Yurii <iuriish@yahoo.com> * - add switch to scalar case execReduceXD functions Signed-off-by: Yurii <iuriish@yahoo.com> * - add support for vectors old shape in ConstantShapeHelper::createShapeInfoWithNoUnitiesForReduce Signed-off-by: Yurii <iuriish@yahoo.com> * - correct cuda mirrorPad Signed-off-by: Yurii <iuriish@yahoo.com> * - add support for vectors old shape in cuda createShapeInfoWithNoUnitiesForReduce Signed-off-by: Yurii <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com> * Add support for CUDA 11.0 (#492) * Add support for CUDA 11.0 * libnd4j tweaks for CUDA 11 Signed-off-by: raver119@gmail.com <raver119@gmail.com> * bindings update, again? Signed-off-by: raver119@gmail.com <raver119@gmail.com> * * Update versions of JavaCPP Presets for FFmpeg, OpenBLAS, and NumPy * update API to match CUDA 8 Signed-off-by: raver119@gmail.com <raver119@gmail.com> * * Update version of JavaCPP Presets for CPython * C++ updated for cuDNN 8.0 Signed-off-by: raver119@gmail.com <raver119@gmail.com> * one more test Signed-off-by: raver119@gmail.com <raver119@gmail.com> * one more test Signed-off-by: raver119@gmail.com <raver119@gmail.com> * one more test Signed-off-by: raver119@gmail.com <raver119@gmail.com> * 128-bit alignment for workspaces Signed-off-by: raver119@gmail.com <raver119@gmail.com> * change seed in 1 test Signed-off-by: raver119@gmail.com <raver119@gmail.com> * Fix dependecy duplication in python4j-parent pom * Fix group id for in python4j-numpy * few tests tweaked Signed-off-by: raver119@gmail.com <raver119@gmail.com> * Remove macosx-x86_64-gpu from nd4j-tests-tensorflow * few minor tweaks for IndexReduce Signed-off-by: raver119@gmail.com <raver119@gmail.com> * one test removed Signed-off-by: raver119@gmail.com <raver119@gmail.com> Co-authored-by: raver119@gmail.com <raver119@gmail.com> Co-authored-by: Serhii Shepel <9946053+sshepel@users.noreply.github.com> * RL4J: Add SyncTrainer and AgentLearnerBuilder for a few algorithms (#504) Signed-off-by: Alexandre Boulanger <aboulang2002@yahoo.com> Co-authored-by: Alexandre Boulanger <44292157+aboulang2002@users.noreply.github.com> Co-authored-by: Yurii Shyrma <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com> Co-authored-by: Serhii Shepel <9946053+sshepel@users.noreply.github.com> * Removed dead code (deeplearning4j#9057) Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * performance improvement (deeplearning4j#9055) * performance improvement Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * revert some changes Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * Development updates (deeplearning4j#9064) * Update versions of JavaCPP Presets for OpenCV, FFmpeg, and MKL Signed-off-by: Samuel Audet <samuel.audet@gmail.com> * Added support for the archunit (deeplearning4j#9062) * Added support for the archunit Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * Updated pom files Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * Datavec code cleaup (deeplearning4j#9071) * removed unnecessary semicolons Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * Use standard charset object Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * Removed unused imports Signed-off-by: Dariusz Zbyrad <dariusz.zbyrad@gmail.com> * WIP: Fix Conv1d causal case * Add inital tests * Update Conv1d tests to be a bit more robust * Remove redundant test * Reset from master * Remove cuda definition (left over) * Update rl4j again * Update pom.xml Co-authored-by: Samuel Audet <samuel.audet@gmail.com> Co-authored-by: Alexandre Boulanger <44292157+aboulang2002@users.noreply.github.com> Co-authored-by: Yurii Shyrma <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com> Co-authored-by: Serhii Shepel <9946053+sshepel@users.noreply.github.com> Co-authored-by: dariuszzbyrad <dariusz.zbyrad@gmail.com>
KonduitAI · Sep 2, 2020 · be4628e · be4628e
1 parent faea1f7
commit be4628e
Show file tree

Hide file tree

Showing 92 changed files with 5,016 additions and 15,914 deletions.
diff --git a/...earning4j-core/src/test/java/org/deeplearning4j/gradientcheck/CNN1DGradientCheckTest.java b/...earning4j-core/src/test/java/org/deeplearning4j/gradientcheck/CNN1DGradientCheckTest.java
@@ -26,6 +26,7 @@
 import org.deeplearning4j.nn.conf.inputs.InputType;
 import org.deeplearning4j.nn.conf.layers.*;
 import org.deeplearning4j.nn.conf.layers.convolutional.Cropping1D;
+import org.deeplearning4j.nn.modelimport.keras.KerasModelImport;
 import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
 import org.deeplearning4j.util.Convolution1DUtils;
 import org.junit.Test;
@@ -38,6 +39,8 @@
 import org.nd4j.linalg.learning.config.NoOp;
 import org.nd4j.linalg.lossfunctions.LossFunctions;
 
+import java.io.File;
+
 import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertTrue;
 
@@ -450,7 +453,7 @@ public void testCnn1dWithMasking(){
     }
 
     @Test
-    public void testCnn1Causal() {
+    public void testCnn1Causal() throws Exception {
         int convNIn = 2;
         int convNOut1 = 3;
         int convNOut2 = 4;
@@ -462,7 +465,6 @@ public void testCnn1Causal() {
         int[] strides   =  {1,   2,  1,  2,  1,  1};
         boolean[] masks =  {false, true, false, true, false, true};
         boolean[] hasB  =  {true, false, true, false, true, true};
-
         for (int i = 0; i < lengths.length; i++) {
             int length = lengths[i];
             int k = kernels[i];
@@ -471,7 +473,7 @@ public void testCnn1Causal() {
             boolean mask = masks[i];
             boolean hasBias = hasB[i];
             //TODO has bias
-            String s = "k=" + k + ", s=" + st + "d=" + d + ", seqLen=" + length;
+            String s = "k=" + k + ", s=" + st + " d=" + d + ", seqLen=" + length;
             log.info("Starting test: " + s);
             Nd4j.getRandom().setSeed(12345);
 

diff --git a/...g4j-core/src/test/java/org/deeplearning4j/nn/layers/convolution/ConvolutionLayerTest.java b/...g4j-core/src/test/java/org/deeplearning4j/nn/layers/convolution/ConvolutionLayerTest.java
diff --git a/...j-modelimport/src/main/java/org/deeplearning4j/nn/modelimport/keras/KerasModelImport.java b/...j-modelimport/src/main/java/org/deeplearning4j/nn/modelimport/keras/KerasModelImport.java
@@ -47,7 +47,7 @@ public class KerasModelImport {
      * @return ComputationGraph
      * @see ComputationGraph
      */
-    public static ComputationGraph importKerasModelAndWeights( InputStream modelHdf5Stream, boolean enforceTrainingConfig)
+    public static ComputationGraph importKerasModelAndWeights(InputStream modelHdf5Stream, boolean enforceTrainingConfig)
             throws IOException, UnsupportedKerasConfigurationException, InvalidKerasConfigurationException{
         File f = null;
         try{

diff --git a/...java/org/deeplearning4j/nn/modelimport/keras/layers/convolutional/KerasConvolution1D.java b/...java/org/deeplearning4j/nn/modelimport/keras/layers/convolutional/KerasConvolution1D.java
@@ -28,6 +28,7 @@
 import org.deeplearning4j.nn.modelimport.keras.exceptions.InvalidKerasConfigurationException;
 import org.deeplearning4j.nn.modelimport.keras.exceptions.UnsupportedKerasConfigurationException;
 import org.deeplearning4j.nn.modelimport.keras.utils.KerasConstraintUtils;
+import org.deeplearning4j.nn.modelimport.keras.utils.KerasLayerUtils;
 import org.deeplearning4j.nn.params.ConvolutionParamInitializer;
 import org.deeplearning4j.nn.weights.IWeightInit;
 import org.nd4j.linalg.api.ndarray.INDArray;
@@ -85,7 +86,6 @@ public KerasConvolution1D(Map<String, Object> layerConfig, boolean enforceTraini
         hasBias = getHasBiasFromConfig(layerConfig, conf);
         numTrainableParams = hasBias ? 2 : 1;
         int[] dilationRate = getDilationRate(layerConfig, 1, conf, false);
-
         LayerConstraint biasConstraint = KerasConstraintUtils.getConstraintsFromConfig(
                 layerConfig, conf.getLAYER_FIELD_B_CONSTRAINT(), conf, kerasMajorVersion);
         LayerConstraint weightConstraint = KerasConstraintUtils.getConstraintsFromConfig(
@@ -101,7 +101,7 @@ public KerasConvolution1D(Map<String, Object> layerConfig, boolean enforceTraini
                 .convolutionMode(getConvolutionModeFromConfig(layerConfig, conf))
                 .kernelSize(getKernelSizeFromConfig(layerConfig, 1,  conf, kerasMajorVersion)[0])
                 .hasBias(hasBias)
-                .stride(getStrideFromConfig(layerConfig, 1, conf)[0]).rnnDataFormat(dimOrder == DimOrder.TENSORFLOW? RNNFormat.NWC: RNNFormat.NCW);
+                .stride(getStrideFromConfig(layerConfig, 1, conf)[0]).rnnDataFormat(dimOrder == DimOrder.TENSORFLOW ? RNNFormat.NCW: RNNFormat.NWC);
         int[] padding = getPaddingFromBorderModeConfig(layerConfig, 1, conf, kerasMajorVersion);
         if (hasBias)
             builder.biasInit(0.0);
@@ -113,6 +113,14 @@ public KerasConvolution1D(Map<String, Object> layerConfig, boolean enforceTraini
             builder.constrainBias(biasConstraint);
         if (weightConstraint != null)
             builder.constrainWeights(weightConstraint);
+        if(inputShape != null) {
+            if(dimOrder == DimOrder.THEANO) {
+                builder.nIn(inputShape[0]);
+            }
+            else {
+                builder.nIn(inputShape[1]);
+            }
+        }
         this.layer = builder.build();
     }
 

diff --git a/...ort/src/test/java/org/deeplearning4j/nn/modelimport/keras/e2e/KerasModelEndToEndTest.java b/...ort/src/test/java/org/deeplearning4j/nn/modelimport/keras/e2e/KerasModelEndToEndTest.java
@@ -605,7 +605,7 @@ public void testMaskingTwoValue() throws Exception {
     }
 
     @Test
-    public void testCausalCon1D() throws Exception {
+    public void testCausalConv1D() throws Exception {
         String[] names = new String[]{
                 "causal_conv1d_k2_s1_d1_cl_model.h5",
                 "causal_conv1d_k2_s1_d2_cl_model.h5",
@@ -635,7 +635,7 @@ public void testCausalCon1D() throws Exception {
     }
 
     @Test
-    public void testCon1D() throws Exception {
+    public void testConv1D() throws Exception {
         String[] names = new String[]{
                 "conv1d_k2_s1_d1_cf_same_model.h5",
                 "conv1d_k2_s1_d1_cf_valid_model.h5",
@@ -697,6 +697,7 @@ public void testCon1D() throws Exception {
         }
     }
 
+
     @Test
     public void testActivationLayers() throws Exception {
         String[] names = new String[]{

diff --git a/...j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/conf/MultiLayerConfiguration.java b/...j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/conf/MultiLayerConfiguration.java
@@ -640,7 +640,7 @@ public MultiLayerConfiguration build() {
 
             if(backpropType == BackpropType.TruncatedBPTT && validateTbpttConfig){
                 //Check for invalid combination - tbptt plus LastTimeStepLayer or
-                for( int i=0; i<confs.size(); i++ ){
+                for( int i = 0; i < confs.size(); i++) {
                     Layer l = confs.get(i).getLayer();
                     if(l instanceof LastTimeStep || l instanceof GlobalPoolingLayer){
                         throw new IllegalStateException("Invalid network configuration detected: Truncated backpropagation through time (TBPTT)" +
@@ -698,7 +698,27 @@ public MultiLayerConfiguration build() {
                     if (inputPreProcessor != null) {
                         currentInputType = inputPreProcessor.getOutputType(currentInputType);
                     }
-                    l.setNIn(currentInputType, false); //Don't override the nIn setting, if it's manually set by the user
+                    if(i > 0) {
+                        Layer layer = confs.get(i - 1).getLayer();
+                        //convolution 1d is an edge case where it has rnn input type but the filters
+                        //should be the output
+                        if(layer instanceof Convolution1DLayer) {
+                            Convolution1DLayer convolution1D = (Convolution1DLayer) layer;
+                            if(l instanceof DenseLayer) {
+                                FeedForwardLayer feedForwardLayer = (FeedForwardLayer) l;
+                                InputType.InputTypeRecurrent recurrent = (InputType.InputTypeRecurrent) inputType;
+                                feedForwardLayer.setNIn(recurrent.getTimeSeriesLength());
+                            }
+                            else
+                                l.setNIn(currentInputType, false); //Don't override the nIn setting, if it's manually set by the user
+                        }
+                        else
+                            l.setNIn(currentInputType, false); //Don't override the nIn setting, if it's manually set by the user
+
+                    }
+                    else
+                         l.setNIn(currentInputType, false); //Don't override the nIn setting, if it's manually set by the user
+
 
                     currentInputType = l.getOutputType(i, currentInputType);
                 }

diff --git a/...rning4j-nn/src/main/java/org/deeplearning4j/nn/layers/convolution/Convolution1DLayer.java b/...rning4j-nn/src/main/java/org/deeplearning4j/nn/layers/convolution/Convolution1DLayer.java
@@ -188,7 +188,7 @@ protected Pair<INDArray,INDArray> preOutput(boolean training, boolean forBackpro
         return preOutput;
     }
 
-    protected Pair<INDArray,INDArray> causalConv1dForward(){
+    protected Pair<INDArray,INDArray> causalConv1dForward() {
         //TODO eventually we'll use this for all convolution modes - but only after libnd4j has cuDNN support
         org.deeplearning4j.nn.conf.layers.Convolution1DLayer c = (org.deeplearning4j.nn.conf.layers.Convolution1DLayer) layerConf();
         Conv1DConfig conf = Conv1DConfig.builder()
@@ -197,7 +197,7 @@ protected Pair<INDArray,INDArray> causalConv1dForward(){
                 .d(c.getDilation()[0])
                 .p(c.getPadding()[0])
                 .dataFormat((((org.deeplearning4j.nn.conf.layers.Convolution1DLayer)
-                        layerConf()).getRnnDataFormat()== RNNFormat.NCW)?Conv1DConfig.NCW: Conv1DConfig.NCW)
+                        layerConf()).getRnnDataFormat() == RNNFormat.NCW) ? Conv1DConfig.NCW: Conv1DConfig.NWC)
                 .paddingMode(PaddingMode.CAUSAL)
                 .build();
         INDArray w = getParam(ConvolutionParamInitializer.WEIGHT_KEY);

diff --git a/jumpy/pom.xml b/jumpy/pom.xml
@@ -181,17 +181,4 @@
         </plugins>
     </build>
 
-    <profiles>
-        <profile>
-            <id>nd4j-backend</id>
-            <activation>
-                <property>
-                    <name>libnd4j.cuda</name>
-                </property>
-            </activation>
-            <properties>
-                <nd4j.backend>nd4j-cuda-${libnd4j.cuda}</nd4j.backend>
-            </properties>
-        </profile>
-    </profiles>
 </project>
diff --git a/libnd4j/README.md b/libnd4j/README.md
@@ -208,3 +208,19 @@ To run tests using CUDA backend it's pretty much similar process:
     2. ./blasbuild/cuda/tests_cpu/layers_tests/runtests (.exe on Windows)
 
 
+## Development
+
+In order to extend and update libnd4j, understanding libnd4j's various
+cmake flags is the key. Many of them are in buildnativeoperations.sh.
+The pom.xml is used to integrate and auto configure the project
+for building with deeplearning4j.
+
+At a minimum, you will want to enable tests. An example default set of flags
+for running tests and getting cpu builds working is as follows:
+```bash
+-DSD_CPU=true -DBLAS=TRUE -DSD_ARCH=x86-64 -DSD_EXTENSION= -DSD_LIBRARY_NAME=nd4jcpu -DSD_CHECK_VECTORIZATION=OFF  -DSD_SHARED_LIB=ON -DSD_STATIC_LIB=OFF -DSD_BUILD_MINIFIER=false -DSD_ALL_OPS=true -DCMAKE_BUILD_TYPE=Release -DPACKAGING=none  -DSD_BUILD_TESTS=OFF -DCOMPUTE=all -DOPENBLAS_PATH=C:/Users/agibs/.javacpp/cache/openblas-0.3.10-1.5.4-SNAPSHOT-windows-x86_64.jar/org/bytedeco/openblas/windows-x86_64 -DDEV=FALSE -DCMAKE_NEED_RESPONSE=YES -DMKL_MULTI_THREADED=TRUE -DSD_BUILD_TESTS=YES
+```
+
+The way the main build script works, it dynamically generates a set of flags
+suitable for use for building the projects. Understanding the build script
+will go a long way in to configuring cmake for your particular IDE.
diff --git a/libnd4j/development.md b/libnd4j/development.md