All Downloads are FREE. Search and download functionalities are using the official Maven repository.
  • Log In
  • Register

    • com.simiacryptus.mindseye.layers.MonitoringWrapperLayer Maven / Gradle / Ivy

    There is a newer version: 2.1.0
    Show newest version
    /*
 * Copyright (c) 2019 by Andrew Charneski.
 *
 * The author licenses this file to you 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 com.simiacryptus.mindseye.layers;

import com.google.gson.JsonObject;
import com.simiacryptus.lang.TimedResult;
import com.simiacryptus.lang.UncheckedRunnable;
import com.simiacryptus.lang.UncheckedSupplier;
import com.simiacryptus.mindseye.lang.*;
import com.simiacryptus.ref.lang.RefUtil;
import com.simiacryptus.ref.wrappers.RefArrays;
import com.simiacryptus.ref.wrappers.RefList;
import com.simiacryptus.util.MonitoredItem;
import com.simiacryptus.util.MonitoredObject;
import com.simiacryptus.util.data.PercentileStatistics;
import com.simiacryptus.util.data.ScalarStatistics;

import javax.annotation.Nonnull;
import javax.annotation.Nullable;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;
import java.util.UUID;
import java.util.concurrent.atomic.AtomicLong;

@SuppressWarnings({"serial", "FieldCanBeLocal"})
public final class MonitoringWrapperLayer extends WrapperLayer implements MonitoredItem {

  private final PercentileStatistics backwardPerformance = new PercentileStatistics();
  private final ScalarStatistics backwardSignal = new PercentileStatistics();
  private final PercentileStatistics forwardPerformance = new PercentileStatistics();
  private final ScalarStatistics forwardSignal = new PercentileStatistics();
  private final boolean verbose = false;
  private boolean recordSignalMetrics = false;
  private int totalBatches = 0;
  private int totalItems = 0;

  protected MonitoringWrapperLayer(@Nonnull final JsonObject json, Map rs) {
    super(json, rs);
    if (json.has("forwardPerf")) {
      forwardPerformance.readJson(json.getAsJsonObject("forwardPerf"));
    }
    if (json.has("backwardPerf")) {
      backwardPerformance.readJson(json.getAsJsonObject("backwardPerf"));
    }
    if (json.has("backpropStatistics")) {
      backwardSignal.readJson(json.getAsJsonObject("backpropStatistics"));
    }
    if (json.has("outputStatistics")) {
      forwardSignal.readJson(json.getAsJsonObject("outputStatistics"));
    }
    recordSignalMetrics = json.get("recordSignalMetrics").getAsBoolean();
    totalBatches = json.get("totalBatches").getAsInt();
    totalItems = json.get("totalItems").getAsInt();
  }

  public MonitoringWrapperLayer(final Layer inner) {
    super(inner);
  }

  @Nonnull
  public PercentileStatistics getBackwardPerformance() {
    return backwardPerformance;
  }

  @Nonnull
  public ScalarStatistics getBackwardSignal() {
    return backwardSignal;
  }

  @Nonnull
  public PercentileStatistics getForwardPerformance() {
    return forwardPerformance;
  }

  @Nonnull
  public ScalarStatistics getForwardSignal() {
    return forwardSignal;
  }

  @Nonnull
  @Override
  public Map getMetrics() {
    @Nonnull final Map map = new HashMap<>();
    map.put("class", inner.getClass().getName());
    map.put("totalBatches", totalBatches);
    map.put("totalItems", totalItems);
    map.put("outputStatistics", forwardSignal.getMetrics());
    map.put("backpropStatistics", backwardSignal.getMetrics());
    if (verbose) {
      map.put("forwardPerformance", forwardPerformance.getMetrics());
      map.put("backwardPerformance", backwardPerformance.getMetrics());
    }
    final double batchesPerItem = totalBatches * 1.0 / totalItems;
    map.put("avgMsPerItem", 1000 * batchesPerItem * forwardPerformance.getMean());
    map.put("medianMsPerItem", 1000 * batchesPerItem * forwardPerformance.getPercentile(0.5));
    final double backpropMean = backwardPerformance.getMean();
    final double backpropMedian = backwardPerformance.getPercentile(0.5);
    map.put("avgMsPerItem_Backward", 1000 * batchesPerItem * backpropMean);
    map.put("medianMsPerItem_Backward", 1000 * batchesPerItem * backpropMedian);
    @Nullable final RefList state = state();
    @Nonnull final ScalarStatistics statistics = new PercentileStatistics();
    assert state != null;
    state.stream().flatMapToDouble(Arrays::stream).forEach(statistics::add);
    if (statistics.getCount() > 0) {
      @Nonnull final Map weightStats = new HashMap<>();
      weightStats.put("buffers", state.size());
      weightStats.putAll(statistics.getMetrics());
      map.put("weights", weightStats);
    }
    state.freeRef();
    return map;
  }

  @Nullable
  @Override
  public String getName() {
    return inner.getName();
  }

  @Override
  public void setName(final String name) {
    if (null != inner) {
      inner.setName(name);
    }
  }

  @Nonnull
  @SuppressWarnings("unused")
  public static MonitoringWrapperLayer fromJson(@Nonnull final JsonObject json, Map rs) {
    return new MonitoringWrapperLayer(json, rs);
  }

  @Nonnull
  public MonitoringWrapperLayer addTo2(@Nonnull final MonitoredObject obj) {
    addTo(obj, inner.getName());
    return this.addRef();
  }

  public void addTo(@Nonnull MonitoredObject obj, String name) {
    setName(name);
    obj.addObj(getName(), this.addRef());
    obj.freeRef();
  }

  @Nonnull
  @Override
  public Result eval(@Nonnull final Result... inObj) {
    @Nonnull final AtomicLong passbackNanos = new AtomicLong(0);
    final Result[] wrappedInput = RefArrays.stream(RefUtil.addRef(inObj)).map(result -> {
      boolean alive = result.isAlive();
      TensorList data = result.getData();
      Result.Accumulator accumulator = new Accumulator(passbackNanos, result.getAccumulator());
      result.freeRef();
      return new Result(data, accumulator, alive);
    }).toArray(Result[]::new);
    @Nonnull
    TimedResult timedResult = TimedResult.time(RefUtil.wrapInterface((UncheckedSupplier) () -> {
      return inner.eval(RefUtil.addRef(wrappedInput));
    }, wrappedInput));
    final Result output = timedResult.getResult();
    forwardPerformance.add(timedResult.timeNanos / 1000000000.0);
    timedResult.freeRef();
    totalBatches++;
    final int items = RefArrays.stream(inObj).mapToInt(x -> {
      TensorList temp_31_0009 = x.getData();
      int temp_31_0003 = temp_31_0009.length();
      temp_31_0009.freeRef();
      x.freeRef();
      return temp_31_0003;
    }).max().orElse(1);
    totalItems += items;
    if (recordSignalMetrics) {
      forwardSignal.clear();
      TensorList temp_31_0010 = output.getData();
      temp_31_0010.stream().parallel().forEach(t -> {
        forwardSignal.add(t.getData());
        t.freeRef();
      });
      temp_31_0010.freeRef();
    }
    boolean alive = output.isAlive();
    TensorList data = output.getData();
    Result.Accumulator accumulator = new Accumulator2(passbackNanos, items, output.getAccumulator());
    output.freeRef();
    return new Result(data, accumulator, alive);
  }

  @Nonnull
  @Override
  public JsonObject getJson(Map resources, DataSerializer dataSerializer) {
    @Nonnull final JsonObject json = super.getJson(resources, dataSerializer);
    //json.fn("forwardPerf",forwardPerf.getJson());
    //json.fn("backwardPerf",backwardPerf.getJson());
    json.addProperty("totalBatches", totalBatches);
    json.addProperty("totalItems", totalItems);
    json.addProperty("recordSignalMetrics", recordSignalMetrics);
    return json;
  }

  public void shouldRecordSignalMetrics(boolean recordSignalMetrics) {
    this.recordSignalMetrics = recordSignalMetrics;
  }

  public @SuppressWarnings("unused")
  void _free() {
    super._free();
  }

  @Nonnull
  public @Override
  @SuppressWarnings("unused")
  MonitoringWrapperLayer addRef() {
    return (MonitoringWrapperLayer) super.addRef();
  }

  private static class Accumulator extends Result.Accumulator {

    private final AtomicLong passbackNanos;
    private Result.Accumulator accumulator;

    public Accumulator(AtomicLong passbackNanos, Result.Accumulator accumulator) {
      this.passbackNanos = passbackNanos;
      this.accumulator = accumulator;
    }

    @Override
    public void accept(@Nullable DeltaSet buffer, @Nullable TensorList data) {
      TimedResult timedResult = TimedResult.time(RefUtil.wrapInterface(
          (UncheckedRunnable) () -> {
            this.accumulator.accept(buffer.addRef(), data.addRef());
          }, data, buffer));
      passbackNanos.addAndGet(timedResult.timeNanos);
      timedResult.freeRef();
    }

    public @SuppressWarnings("unused")
    void _free() {
      super._free();
      accumulator.freeRef();
    }
  }

  private class Accumulator2 extends Result.Accumulator {

    private final AtomicLong passbackNanos;
    private final int items;
    private Result.Accumulator accumulator;

    public Accumulator2(AtomicLong passbackNanos, int items, Result.Accumulator accumulator) {
      this.passbackNanos = passbackNanos;
      this.items = items;
      this.accumulator = accumulator;
    }

    @Override
    public void accept(@Nullable DeltaSet buffer, @Nonnull TensorList data) {
      if (recordSignalMetrics) {
        backwardSignal.clear();
        data.stream().parallel().forEach(t -> {
          backwardSignal.add(t.getData());
          t.freeRef();
        });
      }
      TimedResult timedResult1 = TimedResult
          .time(RefUtil.wrapInterface(
              (UncheckedRunnable) () -> {
                DeltaSet buffer1 = buffer == null ? null : buffer.addRef();
                Result.Accumulator accumulator = this.accumulator;
                try {
                  accumulator.accept(buffer1, data.addRef());
                } finally {
                  accumulator.freeRef();
                }
              },
              data, accumulator.addRef(),
              buffer));
      backwardPerformance.add((timedResult1.timeNanos
          - passbackNanos.getAndSet(0)) / (items * 1e9));
      timedResult1.freeRef();
    }

    public @SuppressWarnings("unused")
    void _free() {
      super._free();
      accumulator.freeRef();
    }
  }
}
    
    
    
    

    




    
    
    © 2015 - 2025 Weber Informatics LLC | Privacy Policy