How to compute predict output according to the tree models exported before？

In regression, the output isn’t equal to the sum of every tree.

Like this,

params = {

‘booster’: ‘gbtree’,

‘objective’: ‘reg:gamma’, # 回归的损失函数，gmma回归

‘gamma’: 0.1,

‘max_depth’: 5,

‘lambda’: 2,

‘subsample’: 0.7,

‘colsample_bytree’: 0.7,

‘min_child_weight’: 3,

‘silent’: 1,

‘eta’: 0.1,

‘seed’: 1000,

‘nthread’: 4,

}

plst = params.items()

and the tree models are as following:

booster[0]:

0:leaf=0.0977716669

booster[1]:

0:leaf=0.0975570381

booster[2]:

0:leaf=0.0972924531

booster[3]:

0:leaf=0.0970089585

and the sum is about 0.4 but the prediction is about 0.7?

should I use eta when compute the sum?