How to use the @tensorflow/tfjs-node.tensor function in @tensorflow/tfjs-node

function getData() {
    var set = mnist.set(8000, 0);

    var trainingSet = set.training;
    var data = []

    for (var i = 0; i < trainingSet.length; i++) {
        data.push(trainingSet[i].input);
        data[i].push(trainingSet[i].output.indexOf(1));
    }
    return tf.tensor(data).as2D(8000, 785);
}

let start = new Date()
      let {
        tensorData,
        tensorLabels,
        vocabulary,
        vocabularyLength
      } = preprocessData(devIntents, params)
      let embeddingMatrix = await getEmbeddingMatrix({
        vocabulary,
        vocabularyLength,
        params
      })
      this.models[params.language] = embeddingLSTMModel({
        params,
        vocabularyLength,
        embeddingMatrix: tf.tensor(embeddingMatrix),
        outputDim: Object.keys(devIntents.intentsDict).length
      })
      this.models[params.language].summary()
      this.models[params.language].compile({
        optimizer: tf.train.adam(params.LEARNING_RATE),
        loss: 'categoricalCrossentropy',
        metrics: ['accuracy']
      })
      console.log('TRAINING...')

      const history = await this.models[params.language].fit(
        tensorData,
        tensorLabels,
        {
          epochs: params.EPOCHS,
          validationSplit: params.VALIDATION_SPLIT

DMLDB.db.get(dml_request.repo, function(err, doc) {
            if (err) { return console.log(err); }
            var data = tfjs_1.tensor(doc.data).as2D(doc.rows, doc.cols);
            if (dml_request.action == 'TRAIN') {
                if (!(dml_request.id in doc.sessions)) {
                    doc.sessions[dml_request.id] = 0;
                    DMLDB.db.put(doc);
                }
                var session_round = doc.sessions[dml_request.id];
                if (session_round+ 1 != dml_request.round) {
                    console.log("Ignoring server's message...");
                    console.log("Request's round was " + dml_request.round + " and current round is " + session_round);
                    return;
                }
            }
            callback(data, dml_request, model);
        });
    }

Object.keys(experiences).map(function(key) {
        tensorified[key] = tf.tensor(experiences[key]);        
      });      
      const {states, actions, rewards, nextStates, dones} = tensorified;

static _labelData(data, label_index) {
        if (label_index < 0) {
            label_index = data[0].length - 1;
        }
        var trainXs = data;
        var trainYs = trainXs.map(function (row) { return row[label_index]; });
        trainXs.forEach(function (x) { x.splice(label_index, 1); });
        return [tfjs_1.tensor(trainXs), tfjs_1.tensor(trainYs)];
    };