Releasing 0.3.3

Author: karllessard

README.md

@@ -61,12 +61,12 @@ systems, you should add the following dependencies:
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
 </dependency>
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
   <classifier>linux-x86_64${javacpp.platform.extension}</classifier>
 </dependency>
 ```
@@ -77,24 +77,24 @@ native dependencies as follows:
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
 </dependency>
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
   <classifier>linux-x86_64${javacpp.platform.extension}</classifier>
 </dependency>
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
   <classifier>macosx-x86_64${javacpp.platform.extension}</classifier>
 </dependency>
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-api</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
   <classifier>windows-x86_64${javacpp.platform.extension}</classifier>
 </dependency>
 ```
@@ -107,7 +107,7 @@ artifact includes transitively all the artifacts above as a single dependency:
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-core-platform${javacpp.platform.extension}</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
 </dependency>
 ```
 
@@ -152,6 +152,7 @@ This table shows the mapping between different version of TensorFlow for Java an
 | 0.3.0  | 2.4.1  |
 | 0.3.1  | 2.4.1  |
 | 0.3.2  | 2.4.1  |
+| 0.3.3  | 2.4.1  |
 
 ## How to Contribute?
 

ndarray/README.md

@@ -11,7 +11,7 @@ To import the NdArray library in your project, simply add the following dependen
 <dependency>
   <groupId>org.tensorflow</groupId>
   <artifactId>ndarray</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
 </dependency>
 ```
 

ndarray/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.tensorflow</groupId>
     <artifactId>tensorflow-java</artifactId>
-    <version>0.3.2</version>
+    <version>0.3.3</version>
   </parent>
   <artifactId>ndarray</artifactId>
   <packaging>jar</packaging>

pom.xml

@@ -5,7 +5,7 @@
 
   <groupId>org.tensorflow</groupId>
   <artifactId>tensorflow-java</artifactId>
-  <version>0.3.2</version>
+  <version>0.3.3</version>
   <packaging>pom</packaging>
 
   <name>TensorFlow Java Parent</name>

tensorflow-core/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.tensorflow</groupId>
     <artifactId>tensorflow-java</artifactId>
-    <version>0.3.2</version>
+    <version>0.3.3</version>
   </parent>
   <artifactId>tensorflow-core</artifactId>
   <packaging>pom</packaging>

tensorflow-core/tensorflow-core-api/pom.xml

@@ -6,7 +6,7 @@
   <parent>
     <groupId>org.tensorflow</groupId>
     <artifactId>tensorflow-core</artifactId>
-    <version>0.3.2</version>
+    <version>0.3.3</version>
   </parent>
   <artifactId>tensorflow-core-api</artifactId>
   <packaging>jar</packaging>

tensorflow-core/tensorflow-core-api/pom.xml.asc

@@ -453,19 +453,18 @@ public Output<?>[] whileLoop(
   synchronized SaverDef saverDef() {
     if (saverDef == null) {
       // Check to see if this graph has a restore operation
-      if (operation("save/restore_all") == null) {
+      if (operation(SAVER_DEF_SCOPE + "/" + SAVER_DEF_RESTORE_OP) == null) {
         // No saver, create one by mutating the graph
         saverDef = addVariableSaver(this);
       } else {
         // This graph already has saving/restoring operations,
-        // regenerate SaverDef without mutating. The names mirror
-        // the python implementation for compatibility.
-        // https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/training/saver.py
-        saverDef = SaverDef.newBuilder()
-                           .setFilenameTensorName("save/filename")
-                           .setSaveTensorName("save/control_dependency")
-                           .setRestoreOpName("save/restore_all")
-                           .build();
+        // regenerate SaverDef without mutating.
+        saverDef =
+            SaverDef.newBuilder()
+                .setFilenameTensorName(SAVER_DEF_SCOPE + "/" + SAVER_DEF_FILENAME_OP + ":0")
+                .setSaveTensorName(SAVER_DEF_SCOPE + "/" + SAVER_DEF_SAVE_OP)
+                .setRestoreOpName(SAVER_DEF_SCOPE + "/" + SAVER_DEF_RESTORE_OP)
+                .build();
       }
     }
     return saverDef;
@@ -570,6 +569,13 @@ public void remove() {
     private int position;
   }
 
+  // These names mirror the python implementation, to reduce the risk of incompatibility.
+  // https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/training/saver.py
+  private static final String SAVER_DEF_SCOPE = "save";
+  private static final String SAVER_DEF_FILENAME_OP = "filename";
+  private static final String SAVER_DEF_SAVE_OP = "control_dependency";
+  private static final String SAVER_DEF_RESTORE_OP = "restore_all";
+
   private static TF_Graph allocate() {
       return TF_NewGraph();
   }
@@ -797,7 +803,7 @@ private static Object[] whileLoop(
   }
 
   private static SaverDef addVariableSaver(Graph graph) {
-    Ops tf = Ops.create(graph).withSubScope("save");
+    Ops tf = Ops.create(graph).withSubScope(SAVER_DEF_SCOPE);
 
     List<String> varNames = new ArrayList<>();
     List<Operand<?>> varOutputs = new ArrayList<>();
@@ -812,36 +818,35 @@ private static SaverDef addVariableSaver(Graph graph) {
       }
     }
 
-    // FIXME Need an easier way to initialize an NdArray from a list
-    String[] tmp = new String[varNames.size()];
-    Constant<TString> varNamesTensor = tf.constant(StdArrays.ndCopyOf(varNames.toArray(tmp)));
-    Operand<TString> varSlices = tf.zerosLike(varNamesTensor);
-
-    Placeholder<TString> saveFilename = tf.withName("filename").placeholder(TString.class);
-    Save saveVariables = tf.train.save(
-        saveFilename,
-        varNamesTensor,
-        varSlices,
-        varOutputs
-    );
-    Identity<TString> id = tf.withControlDependencies(Arrays.asList(saveFilename,saveVariables))
-            .withName("control_dependency").identity(saveFilename);
-    Restore restoreVariables = tf.train.restore(
-        saveFilename,
-        varNamesTensor,
-        varSlices,
-        varTypes
-    );
-    List<Op> restoreOps = new ArrayList<>(varOutputs.size());
-    for (int i = 0; i < varOutputs.size(); ++i) {
-      restoreOps.add(tf.assign(varOutputs.get(i), (Operand) restoreVariables.tensors().get(i)));
+    Placeholder<TString> filename = tf.withName(SAVER_DEF_FILENAME_OP).placeholder(TString.class);
+    Identity<TString> save = null;
+    NoOp restore = null;
+
+    if (varNames.isEmpty()) {
+      save = tf.withName(SAVER_DEF_SAVE_OP).identity(filename);
+      restore = tf.withName(SAVER_DEF_RESTORE_OP).noOp();
+    } else {
+      String[] tmp = new String[varNames.size()];
+      Constant<TString> varNamesTensor = tf.constant(StdArrays.ndCopyOf(varNames.toArray(tmp)));
+      Operand<TString> varSlices = tf.zerosLike(varNamesTensor);
+      Save saveVars = tf.train.save(filename, varNamesTensor, varSlices, varOutputs);
+      List<Op> saveDeps = Arrays.asList(filename, saveVars);
+      Restore restoreVars = tf.train.restore(filename, varNamesTensor, varSlices, varTypes);
+      List<Op> restoreDeps = new ArrayList<>(varOutputs.size());
+      for (int i = 0; i < varOutputs.size(); ++i) {
+        restoreDeps.add(tf.assign(varOutputs.get(i), (Operand) restoreVars.tensors().get(i)));
+      }
+      save = tf.withControlDependencies(saveDeps).withName(SAVER_DEF_SAVE_OP).identity(filename);
+      restore = tf.withControlDependencies(restoreDeps).withName(SAVER_DEF_RESTORE_OP).noOp();
     }
-    NoOp restoreAll = tf.withControlDependencies(restoreOps).withName("restore_all").noOp();
 
+    // 'Filename' must be the name of a tensor (i.e. with output index)
+    // 'Save' must be an operation name, even if the field name is confusing (see SaverDef doc)
+    // 'Restore' must be an operation name
     return SaverDef.newBuilder()
-        .setFilenameTensorName(saveFilename.op().name())
-        .setSaveTensorName(id.op().name())
-        .setRestoreOpName(restoreAll.op().name())
+        .setFilenameTensorName(filename.output().name())
+        .setSaveTensorName(save.op().name())
+        .setRestoreOpName(restore.op().name())
         .build();
   }
 

tensorflow-core/tensorflow-core-api/src/main/java/org/tensorflow/Graph.java

@@ -38,6 +38,11 @@ public int index() {
     return index;
   }
 
+  /** Returns the full name of this Output (a.k.a. tensor name) */
+  public String name() {
+    return op().name() + ":" + index;
+  }
+
   /** Returns the DataType of the tensor referred to by this Output. */
   @SuppressWarnings("unchecked")
   public DataType dataType() {
@@ -48,7 +53,7 @@ public DataType dataType() {
   @SuppressWarnings("unchecked")
   @Override
   public Class<T> type() {
-    return (Class<T>)TensorTypeRegistry.find(dataType()).type();
+    return (Class<T>) TensorTypeRegistry.find(dataType()).type();
   }
 
   /**
@@ -63,7 +68,10 @@ public Class<T> type() {
   public <U extends TType> Output<U> expect(Class<U> type) {
     if (type != type()) {
       throw new IllegalArgumentException(
-          "Cannot cast from output of " + this.type().getSimpleName() + " to output of " + type.getSimpleName());
+          "Cannot cast from output of "
+              + this.type().getSimpleName()
+              + " to output of "
+              + type.getSimpleName());
     }
     return ((Output<U>) this);
   }
@@ -80,17 +88,16 @@ public <U extends TType> Output<U> expect(Class<U> type) {
    *
    * @return tensor
    * @throws IllegalStateException if this output results from a graph
-   * @throws ClassCastException if the type of the tensor and this output are unexpectedly incompatible
+   * @throws ClassCastException if the type of the tensor and this output are unexpectedly
+   *     incompatible
    * @see EagerSession
    */
   @SuppressWarnings("unchecked")
   public T asTensor() {
-    return (T)operation.tensor(index);
+    return (T) operation.tensor(index);
   }
 
-  /**
-   * Returns the (possibly partially known) shape of the tensor referred to by this output.
-   */
+  /** Returns the (possibly partially known) shape of the tensor referred to by this output. */
   @Override
   public Shape shape() {
     return operation.shape(index);

tensorflow-core/tensorflow-core-api/src/main/java/org/tensorflow/Output.java

@@ -29,6 +29,7 @@
 import java.util.LinkedHashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.Objects;
 import java.util.stream.Collectors;
 import org.bytedeco.javacpp.BytePointer;
 import org.bytedeco.javacpp.PointerScope;
@@ -432,7 +433,7 @@ private static SavedModelBundle load(
   }
 
   private static void validateTags(String[] tags) {
-    if (tags == null || Arrays.stream(tags).anyMatch(t -> t == null || t.isEmpty())) {
+    if (tags == null || Arrays.stream(tags).anyMatch(Objects::isNull)) {
       throw new IllegalArgumentException("Invalid tags: " + Arrays.toString(tags));
     }
   }