_train()

in webassembly/fasttext.js [80:118]

• 36 lines of code
• 5 McCabe index (conditional complexity)

  _train(url, modelName, kwargs = {}, callback = null) {
    const fetchFunc = (thisModule && thisModule.fetch) || fetch;
    const fastTextNative = this.f;

    return new Promise(function(resolve, reject) {
      fetchFunc(url).then(response => {
        return response.arrayBuffer();
      }).then(bytes => {
        const byteArray = new Uint8Array(bytes);
        const FS = fastTextModule.FS;
        FS.writeFile(trainFileInWasmFs, byteArray);
      }).then(() =>  {
        const argsList = ['lr', 'lrUpdateRate', 'dim', 'ws', 'epoch',
          'minCount', 'minCountLabel', 'neg', 'wordNgrams', 'loss',
          'model', 'bucket', 'minn', 'maxn', 't', 'label', 'verbose',
          'pretrainedVectors', 'saveOutput', 'seed', 'qout', 'retrain',
          'qnorm', 'cutoff', 'dsub', 'qnorm', 'autotuneValidationFile',
          'autotuneMetric', 'autotunePredictions', 'autotuneDuration',
          'autotuneModelSize'];
        const args = new fastTextModule.Args();
        argsList.forEach(k => {
          if (k in kwargs) {
            args[k] = kwargs[k];
          }
        });
        args.model = fastTextModule.ModelName[modelName];
        args.loss = ('loss' in kwargs) ?
          fastTextModule.LossName[kwargs['loss']] : 'hs';
        args.thread = 1;
        args.input = trainFileInWasmFs;

        fastTextNative.train(args, callback);

        resolve(new FastTextModel(fastTextNative));
      }).catch(error => {
        reject(error);
      });
    });
  }