Created by: 333caowei

input数据如下：一个是id、一个是id list

query = fluid.layers.data(
                    name="query",
                    shape=[1],
                    dtype='int64')

query_char = fluid.layers.data(
                    name="query_char",
                    shape=[1],
                    dtype='int64',
                    lod_level=1)

然后使用下面的save_inference_model保存模型

fluid.io.save_inference_model(epoch_model, ["query", "query_char"], predict, exe)

载入模型：

[inference_program, feed_target_names, fetch_targets] = fluid.io.load_inference_model(model_path, exe)

预测时候遇到问题：

score = exe.run(inference_program,
                feed=data2tensor({"query": 4, "query_char": [3,2,5]}, place),
                fetch_list=fetch_targets,
                return_numpy=True)

def data2tensor(data, place):
    res = dict()
    res["query"] = to_lodtensor(data["query"], place)
    res["query_char"] = to_lodtensor(data["query_char"], place)
    return res

def to_lodtensor(data, place):
    res = fluid.LoDTensor()
    res.set(data, place)
    return res

在预测时候，data2tensor函数调用to_lodtensor将数据转化后给到模型,报如下错误： paddle.fluid.core.EnforceNotMet: enforce ids_dims.size() == 2 failed, 0 != 2 at [/paddle/paddle/fluid/operators/lookup_table_op.cc:43] PaddlePaddle Call Stacks: