params.py

    def get_model_params(self):
        # get trainable params.
        model_names = []
        model_params = []
        model_shapes = []
        with self.g.as_default():
            t_vars = tf.trainable_variables()
            for var in t_vars:
                param_name = var.name
                p = self.sess.run(var)
                model_names.append(param_name)
                params = np.round(p*10000).astype(np.int).tolist() # ..?!
                model_params.append(params)
                model_shapes.append(p.shape)
        return model_params, model_shapes, model_names

    def set_model_params(self, params):
        with self.g.as_default():
            trainable_vars = tf.trainable_variables()
            idx = 0
            for var in trainable_vars:
                t_shape = self.sess.run(var).shape
                p_ = np.array(params[idx])
                assert t_shape == p_.shape, "inconsistent shape"
                assign_op = var.assign(p_.astype(np.float)/10000.)
                self.sess.run(assign_op)
                idx += 1
        
    def save_json(self, jsonfile='rnn.json'):
        model_params, model_shapes, model_names = self.get_model_params()
        qparams = []
        for p in model_params:
            qparams.append(p)
        with open(jsonfile, 'wt') as outfile:
            json.dump(qparams, outfile, sort_keys=True, indent=0, separators=(',', ': '))

    def load_json(self, jsonfile='rnn.json'):
        with open(jsonfile, 'r') as f:
            params = json.load(f)
        self.set_model_params(params)