GO TO THE SECOND BOX AND TRAIN THE MODEL IF THIS IS YOUR FIRST TIME.
This is the xOr model with pre-saved weights loaded and used to predict the outcome. Expect close to 0, 1, 1, 0 for outputs. Took 138 seconds to train 10,000 batches (epoch only 1) on my browser

{
   "keras_version": "2.1.2",
   "backend": "tensorflow",
   "weights": {
      "hiddenLayer": [
         {
            "name": "denseLayer/kernel:0",
            "dtype": "float32",
            "shape": [
               2,
               3
            ],
            "value": [
               [
                  -7.0955453,
                  4.7261367,
                  -5.5323949
               ],
               [
                  4.9381752,
                  5.4428835,
                  7.3709011
               ]
            ]
         },
         {
            "name": "denseLayer/bias:0",
            "dtype": "float32",
            "shape": [
               3
            ],
            "value": [
               -1.7896707,
               -0.3296011,
               2.4894311
            ]
         }
      ],
      "outputLayer": [
         {
            "name": "denseLayer/kernel:0",
            "dtype": "float32",
            "shape": [
               3,
               1
            ],
            "value": [
               [
                  8.2123137
               ],
               [
                  5.1289654
               ],
               [
                  -8.3957253
               ]
            ]
         },
         {
            "name": "denseLayer/bias:0",
            "dtype": "float32",
            "shape": [
               1
            ],
            "value": [
               -0.6637133
            ]
         }
      ]
   }
}

json-upload: weights-upload:

This can create the weights used in the above example. Look at the console for more information about the Tensors. Increase the batches to get values close to 0, 1, 1, 0 for outputs.

Tensorflowjs version 0.11.0 just made it easierr to load and save models and weights

Batches

Local or Download file name DON'T CHANGE THIS FILE NAME WHEN SAVING TO COMPUTER AS IT NAMES THE WEIGHTS FILE RELATIVE PATH!

View file system saved files

Tensorflowjs Easy weights loading