Callbacks phase 1

tensorflow-framework/pom.xml

@@ -43,6 +43,12 @@
       <groupId>org.tensorflow</groupId>
       <artifactId>tensorflow-core-api</artifactId>
       <version>${project.version}</version>
+    </dependency>
+    <!-- https://mvnrepository.com/artifact/org.apache.commons/commons-csv -->
+    <dependency>
+      <groupId>org.apache.commons</groupId>
+      <artifactId>commons-csv</artifactId>
+      <version>1.8</version>
     </dependency>
       <dependency>
         <groupId>org.junit.jupiter</groupId>

tensorflow-framework/src/main/java/org/tensorflow/framework/callbacks/CSVLogger.java

@@ -0,0 +1,277 @@
+/* Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+=======================================================================*/
+package org.tensorflow.framework.callbacks;
+
+import org.apache.commons.csv.CSVFormat;
+import org.apache.commons.csv.CSVPrinter;
+import org.tensorflow.ndarray.NdArray;
+import org.tensorflow.ndarray.Shape;
+import org.tensorflow.types.family.TNumber;
+
+import java.io.File;
+import java.io.FileWriter;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Map;
+import java.util.StringJoiner;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+import java.util.stream.Collectors;
+
+/**
+ * Callback that streams epoch results to a CSV file.
+ *
+ * <p>Supports all values that can be represented as a string
+ *
+ * @param <T> the data type for the weights in the model
+ */
+public class CSVLogger<T extends TNumber> extends Callback implements AutoCloseable {
+
+  public static final char DEFAULT_SEPARATOR = ',';
+  public static final boolean DEFAULT_APPEND = false;
+
+  private final File file;
+  private final char separator;
+  private final boolean append;
+  private List<String> keys;
+  private boolean appendHeader = true;
+
+  private CSVPrinter writer;
+
+  /**
+   * Creates a CSVLogger callback using {@link #DEFAULT_SEPARATOR} to separate elements in the csv
+   * file, and {@link #DEFAULT_APPEND} for the append value.
+   *
+   * @param file the csv file
+   */
+  public CSVLogger(File file) {
+    this(file, DEFAULT_SEPARATOR, DEFAULT_APPEND);
+  }
+
+  /**
+   * Creates a CSVLogger callback using {@link #DEFAULT_SEPARATOR} to separate elements in the csv
+   * file, and {@link #DEFAULT_APPEND} for the append value.
+   *
+   * @param filename filename of the csv file
+   */
+  public CSVLogger(String filename) {
+    this(new File(filename), DEFAULT_SEPARATOR, DEFAULT_APPEND);
+  }
+
+  /**
+   * Creates a CSVLogger callback using {@link #DEFAULT_APPEND} for the append value.
+   *
+   * @param file the csv file
+   * @param separator string used to separate elements in the csv file.
+   */
+  public CSVLogger(File file, char separator) {
+    this(file, separator, false);
+  }
+
+  /**
+   * Creates a CSVLogger callback using {@link #DEFAULT_APPEND} for the append value.
+   *
+   * @param filename filename of the csv file
+   * @param separator string used to separate elements in the csv file.
+   */
+  public CSVLogger(String filename, char separator) {
+    this(new File(filename), separator, false);
+  }
+
+  /**
+   * Creates a CSVLogger callback.
+   *
+   * @param filename filename of the csv file
+   * @param separator the character used to separate elements in the csv file.
+   * @param append if true, append if file exists (useful for continuing training). if false,
+   *     overwrite existing file.
+   */
+  public CSVLogger(String filename, char separator, boolean append) {
+    this(new File(filename), separator, append);
+  }
+
+  /**
+   * Creates a CSVLogger callback.
+   *
+   * @param file the csv file
+   * @param separator the character used to separate elements in the csv file.
+   * @param append if true, append if file exists (useful for continuing training). if false,
+   *     overwrite existing file.
+   */
+  public CSVLogger(File file, char separator, boolean append) {
+    this.file = file;
+    this.separator = separator;
+    this.append = append;
+  }
+
+  /** {@inheritDoc} */
+  @Override
+  public void onTrainBegin(Map<String, Number> logs) {
+    appendHeader = !append || !file.exists();
+  }
+
+  // TODO Should we handle Java arrays??
+  @SuppressWarnings("unchecked")
+  private String handleValue(Object val) {
+
+    if (val instanceof String) {
+      return val.toString();
+    } else if (val instanceof NdArray) { // todo
+      boolean isScalar = ((NdArray<?>) val).rank() == 0;
+      if (isScalar) {
+        return ((NdArray<?>) val).getObject().toString();
+      } else {
+        NdArray<?> array = (NdArray<T>) val;
+        return ndArrayToString(array);
+      }
+    } else if (val instanceof Collection) {
+      return "["
+          + ((Collection<T>) val).stream().map(Object::toString).collect(Collectors.joining(","))
+          + "]";
+    } else {
+      return val.toString();
+    }
+  }
+
+  /**
+   * coverts an NdArray to a printable string
+   *
+   * @param ndArray the NdArray
+   * @return the printable string
+   */
+  private String ndArrayToString(NdArray<?> ndArray) {
+    Iterator<? extends NdArray<?>> iterator = ndArray.scalars().iterator();
+    Shape shape = ndArray.shape();
+    if (shape.numDimensions() == 0) {
+      if (!iterator.hasNext()) {
+        return "";
+      }
+      return valToString(iterator.next().getObject());
+    }
+    return ndArrayToString(iterator, shape, 0);
+  }
+
+  /**
+   * coverts an NdArray iterator to a printable string
+   *
+   * @param iterator the NdArray iterator
+   * @param shape the shape of the NdArray item
+   * @param dimension the dimension within the overall NDArray tree
+   * @return the printable string
+   */
+  private String ndArrayToString(Iterator<? extends NdArray<?>> iterator, Shape shape, int dimension) {
+    if (dimension < shape.numDimensions() - 1) {
+      StringJoiner joiner = new StringJoiner("", "[", "]");
+      for (long i = 0, size = shape.size(dimension); i < size; ++i) {
+        String element = ndArrayToString(iterator, shape, dimension + 1);
+        joiner.add(element);
+      }
+      return joiner.toString();
+    } else {
+      StringJoiner joiner = new StringJoiner(", ", "[", "]");
+      for (long i = 0, size = shape.size(dimension); i < size; ++i) {
+        Object element = iterator.next().getObject();
+        joiner.add(valToString(element));
+      }
+      return joiner.toString();
+    }
+  }
+
+  /**
+   * Converts a value to a printable string
+   *
+   * @param val the value
+   * @return the printable string
+   */
+  private String valToString(Object val) {
+    if (val instanceof Number) {
+      Number nVal = (Number) val;
+      if (nVal instanceof Float || nVal instanceof Double) {
+        return String.format("%e", nVal.doubleValue());
+      } else if (nVal instanceof Byte) {
+        return String.format("0x%2x", nVal.byteValue());
+      } else {
+        return String.format("%d", nVal.longValue());
+      }
+    } else {
+      return val.toString();
+    }
+  }
+
+  /** {@inheritDoc} */
+  @Override
+  @SuppressWarnings("unchecked")
+  public void onEpochEnd(int epoch, Map<String, Number> logs) {
+    logs = logs == null ? Collections.EMPTY_MAP : logs;
+
+    if (keys == null) {
+      keys = new ArrayList<>(logs.keySet());
+      Collections.sort(this.keys);
+    }
+
+    if (writer == null) {
+      try {
+        List<String> fieldNames = new ArrayList<>();
+        fieldNames.add("epoch");
+        fieldNames.addAll(this.keys);
+        CSVFormat csvFormat =
+            appendHeader
+                ? CSVFormat.EXCEL
+                    .withHeader(fieldNames.toArray(new String[0]))
+                    .withDelimiter(separator)
+                : CSVFormat.EXCEL.withDelimiter(separator);
+        writer = new CSVPrinter(new FileWriter(file, append), csvFormat);
+      } catch (IOException ex) {
+        Logger.getLogger(CSVLogger.class.getName()).log(Level.SEVERE, null, ex);
+        return;
+      }
+    }
+
+    /* TODO include when integrated with Model
+    if (getModel().isStopTraining()) {
+      final Map<String, Number> flogs = logs;
+      keys.forEach(
+          key -> {
+            if (!flogs.containsKey(key)) {
+              flogs.put(key, Double.NaN);
+            }
+          });
+    }
+     */
+    try {
+      final List<String> values = new ArrayList<>();
+      final Map<String, Number> logsFinal = logs;
+      values.add(String.valueOf(epoch));
+      keys.forEach(key -> values.add(handleValue(logsFinal.get(key))));
+      writer.printRecord(values);
+      writer.flush();
+    } catch (IOException ex) {
+      Logger.getLogger(CSVLogger.class.getName()).log(Level.SEVERE, null, ex);
+    }
+  }
+
+  /** {@inheritDoc} */
+  @Override
+  public void close() throws IOException {
+    if (writer != null) {
+      writer.close();
+      writer = null;
+    }
+  }
+}