How to use the @tensorflow/tfjs.div function in @tensorflow/tfjs

const numAnchors = anchors.shape[0];
  // Reshape to height, width, num_anchors, box_params.
  const anchorsTensor = tf.reshape(anchors, [1, 1, numAnchors, 2]);

  const gridShape = feats.shape.slice(1, 3); // height, width

  const gridY = tf.tile(tf.reshape(tf.range(0, gridShape[0]), [-1, 1, 1, 1]), [1, gridShape[1], 1, 1]);
  const gridX = tf.tile(tf.reshape(tf.range(0, gridShape[1]), [1, -1, 1, 1]), [gridShape[0], 1, 1, 1]);
  const grid = tf.concat([gridX, gridY], 3).cast(feats.dtype);

  feats = feats.reshape([gridShape[0], gridShape[1], numAnchors, numClasses + 5]);

  const [xy, wh, con, probs] = tf.split(feats, [2, 2, 1, numClasses], 3);
  // Adjust preditions to each spatial grid point and anchor size.
  const boxXy = tf.div(tf.add(tf.sigmoid(xy), grid), gridShape.reverse());
  const boxWh = tf.div(tf.mul(tf.exp(wh), anchorsTensor), inputShape.reverse());
  const boxConfidence = tf.sigmoid(con);

  let boxClassProbs;
  if (isV3) {
    boxClassProbs = tf.sigmoid(probs);
  } else {
    boxClassProbs = tf.softmax(probs);
  }

  return [boxXy, boxWh, boxConfidence, boxClassProbs];
}

) {
  const numAnchors = anchors.shape[0];
  // Reshape to height, width, num_anchors, box_params.
  const anchorsTensor = tf.reshape(anchors, [1, 1, numAnchors, 2]);

  const gridShape = feats.shape.slice(1, 3); // height, width

  const gridY = tf.tile(tf.reshape(tf.range(0, gridShape[0]), [-1, 1, 1, 1]), [1, gridShape[1], 1, 1]);
  const gridX = tf.tile(tf.reshape(tf.range(0, gridShape[1]), [1, -1, 1, 1]), [gridShape[0], 1, 1, 1]);
  const grid = tf.concat([gridX, gridY], 3).cast(feats.dtype);

  feats = feats.reshape([gridShape[0], gridShape[1], numAnchors, numClasses + 5]);

  const [xy, wh, con, probs] = tf.split(feats, [2, 2, 1, numClasses], 3);
  // Adjust preditions to each spatial grid point and anchor size.
  const boxXy = tf.div(tf.add(tf.sigmoid(xy), grid), gridShape.reverse());
  const boxWh = tf.div(tf.mul(tf.exp(wh), anchorsTensor), inputShape.reverse());
  const boxConfidence = tf.sigmoid(con);

  let boxClassProbs;
  if (isV3) {
    boxClassProbs = tf.sigmoid(probs);
  } else {
    boxClassProbs = tf.softmax(probs);
  }

  return [boxXy, boxWh, boxConfidence, boxClassProbs];
}

const actionIndexTensor = await tf.tidy(() => {
      if (Array.isArray(x)) x = tf.tensor(x);
      let predictions = tf.squeeze(this.predict(x.expandDims(0)));

      if (this.isGreedy) {
        return tf.argMax(predictions).arraySync()[0];
      }

      // Scale to [0, 1]
      predictions = tf.div(tf.add(1, predictions), 2);
      // Due to noise, can still be larger/smaller
      const min = tf.min(predictions);
      const max = tf.max(predictions);
      predictions = tf.div(tf.sub(predictions, min), tf.sub(max, min));

      // Now sample with the given temperature
      const logProbs = tf.div(tf.log(predictions), temperature);
      const expProbs = tf.exp(logProbs);
      const sumExpProbs = tf.sum(expProbs);
      const probs = tf.div(expProbs, sumExpProbs);

      return tf.multinomial(probs, 1, null, true);
    });
    const actionIndex = (await actionIndexTensor.array())[0];

const onehot = onehotBuffer.toTensor();
      let output;
      if (this.model.embedding) {
        const embedded = tf.matMul(onehot, this.model.embedding);
        output = tf.multiRNNCell(this.cells, embedded, this.state.c, this.state.h);
      } else {
        output = tf.multiRNNCell(this.cells, onehot, this.state.c, this.state.h);
      }

      this.state.c = output[0];
      this.state.h = output[1];

      const outputH = this.state.h[1];
      const weightedResult = tf.matMul(outputH, this.model.fullyConnectedWeights);
      const logits = tf.add(weightedResult, this.model.fullyConnectedBiases);
      const divided = tf.div(logits, tf.tensor(temperature));
      const probabilities = tf.exp(divided);
      probabilitiesNormalized = await tf.div(
        probabilities,
        tf.sum(probabilities),
      ).data();

      if (i < userInput.length - 1) {
        input = encodedInput[i + 1];
      } else {
        input = sampleFromDistribution(probabilitiesNormalized);
        results.push(input);
      }
    }

    let generated = '';
    results.forEach((char) => {

const result = tf.tidy(() => {
      return tf.div(tf.add(inputDeproc, tf.scalar(1)), tf.scalar(2));
    });
    inputDeproc.dispose();

function sampleLogits(
  logits: tf.Tensor1D,
  temperature?: number,
  seed?: number) {
  temperature = temperature !== undefined ? temperature : 1.;
  if (temperature < 0. || temperature > 1.) {
    throw new Error('Invalid temperature specified');
  }

  let result: tf.Scalar;

  if (temperature === 0) {
    result = tf.argMax(logits, 0) as tf.Scalar;
  } else {
    if (temperature < 1) {
      logits = tf.div(logits, tf.scalar(temperature, 'float32'));
    }
    const scores = tf.reshape(tf.softmax(logits, 0), [1, -1]) as tf.Tensor2D;
    const sample = tf.multinomial(scores, 1, seed, true) as tf.Tensor1D;
    result = tf.reshape(sample, []) as tf.Scalar;
  }

  return result;
}

export function yolo_boxes_to_corners(box_xy, box_wh) {
    const two = tf.tensor1d([2.0]);
    const box_mins = tf.sub(box_xy, tf.div(box_wh, two));
    const box_maxes = tf.add(box_xy, tf.div(box_wh, two));

    const dim_0 = box_mins.shape[0];
    const dim_1 = box_mins.shape[1];
    const dim_2 = box_mins.shape[2];
    const size = [dim_0, dim_1, dim_2, 1];

    return tf.concat([
        box_mins.slice([0, 0, 0, 1], size),
        box_mins.slice([0, 0, 0, 0], size),
        box_maxes.slice([0, 0, 0, 1], size),
        box_maxes.slice([0, 0, 0, 0], size),
    ], 3);
}

export function yolo_boxes_to_corners(box_xy, box_wh) {
  const two = tf.tensor1d([2.0]);
  const box_mins = tf.sub(box_xy, tf.div(box_wh, two));
  const box_maxes = tf.add(box_xy, tf.div(box_wh, two));

  const dim_0 = box_mins.shape[0];
  const dim_1 = box_mins.shape[1];
  const dim_2 = box_mins.shape[2];
  const size = [dim_0, dim_1, dim_2, 1];

  return tf.concat([
    box_mins.slice([0, 0, 0, 1], size),
    box_mins.slice([0, 0, 0, 0], size),
    box_maxes.slice([0, 0, 0, 1], size),
    box_maxes.slice([0, 0, 0, 0], size),
  ], 3);
}