Run results for: tensorflow-preprocessing

Run details:

User agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/78.0.3904.108 Chrome/78.0.3904.108 Safari/537.36

Browser: Chromium 78

Operating system: Ubuntu

Device Platform: Desktop

Date tested: 4 years ago

Test name	Executions per second
await runPrediction([], [])	730636900.0 Ops/sec
await runPrediction([1], [1])	717877400.0 Ops/sec

HTML Preparation code:

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs/dist/tf.min.js"></script>

Script Preparation code:

function runPrediction(graphsArray, featuresArray)
{
    let input = this.tf.tidy(() =>
    {
      let featuresTensor = null;
      let graphsTensor = null;

      if (graphsArray && graphsArray.length > 0)
      {
        let graphsTensors = [];
        for (let graph of graphsArray)
        {
          let graphTensor = this.tf.tensor(graph);
          let shape = graphTensor.shape;
          if (shape[0] > GRAPH_SIZE)
          {
            graphTensor = graphTensor.slice(
              [0, 0], [GRAPH_SIZE, GRAPH_SIZE]
            );
          }
          else
          {
            graphTensor = graphTensor.pad([
              [0, GRAPH_SIZE - shape[0]],
              [0, GRAPH_SIZE - shape[1]]
            ]);
          }
          graphTensor = this._localPooling(graphTensor);
          graphsTensors.push(graphTensor);
        }
        graphsTensor = this.tf.stack(graphsTensors);
      }
      else
      {
        graphsTensor = this.tf.zeros([1, GRAPH_SIZE, GRAPH_SIZE]);
      }

      if (featuresArray && featuresArray.length > 0)
      {
        let featuresTensors = [];
        for (let features of featuresArray)
        {
          let featuresPadded = null;
          if (features.length < GRAPH_SIZE)
          {
            features = features.concat(new Array(GRAPH_SIZE).fill(0));
          }
          featuresPadded = features.slice(0, GRAPH_SIZE);

          featuresTensors.push(this.tf.oneHot(
            this.tf.tensor(
              featuresPadded,
              [1, GRAPH_SIZE], "int32"
            ),
            NUM_OF_TAGS)
          );
        }
        featuresTensor = this.tf.cast(
          this.tf.concat(featuresTensors), "float32"
        );
      }
      else
      {
        featuresTensor = this.tf.zeros([1, GRAPH_SIZE, NUM_OF_TAGS]);
      }
      return {graphsTensor, featuresTensor};
    });
}

Tests:

await runPrediction([], [])
1
await runPrediction([1], [1])
2

Run results for: tensorflow-preprocessing

await runPrediction([], [])

await runPrediction([1], [1])