## mapPartions def model_pred(partitionData): updatedData = [] for row in partitionData: pred_value = model.value.predict([row[2:]])[0] pred_value = float(round(pred_value,4)) updatedData.append([row[0],row[1],pred_value]) return iter(updatedData) pred = df.rdd.mapPartitions(model_pred).toDF(['p_number','name',"score"]) model 需要广播