562508c5edf9988078f3ca4300e1d6be569a4b8f,python/timeseries/timeseries_customRNN.py,TsRNNCustom,predict,#TsRNNCustom#Any#Any#,141

Before Change

        if self.n_inputs != 1:
            raise ValueError("Currently only supports univariate input per time-step")
        preds = list()
        x = np.zeros(shape=(1, self.n_lag, 1), dtype=np.float32)
        for i in range(self.n_lag):
            x[0, i, 0] = start_ts[i]
        init_state = np.zeros(shape=(1, self.state_size))
        for i in range(n):
            yhat = self.predict_op.eval(feed_dict={self.X: x,
                                                    self.init_state: init_state})
            logger.debug("pred: %d %s" % (i, str(yhat)))
            preds.append(yhat[0, 0])
            for lag in range(self.n_lag-1):
                x[0, lag, 0] = x[0, lag+1, 0]
            x[0, self.n_lag-1, 0] = yhat[0, 0]
        return np.array(preds)

After Change

        
        if self.n_inputs != 1:
            raise ValueError("Currently only supports univariate input per time-step")
        seq = list(np.reshape(start_ts, newshape=(-1,)))
        logger.debug("seq: %s" % str(seq))
        preds = list()
        init_state = np.zeros(shape=(1, self.state_size))
        for i in range(n):
            ts = seq[-self.n_lag:]
            X_batch = np.array(ts).reshape(1, self.n_lag, self.n_inputs)
            yhat = self.predict_op.eval(feed_dict={self.X: X_batch,
                                                   self.init_state: init_state})
            logger.debug("pred: %d %s" % (i, str(yhat)))
            preds.append(yhat[0, 0])

Instances