ValueError:值张量（“normalize_element/component_0:0”，shape=（），dtype=int32，device=/device:CPU:0）的列组不足以进行批处理

Question

我试图用估计器进行预测，但我得到了以下错误：

ValueError: Value Tensor("normalize_element/component_0:0", shape=(), dtype=int32, device=/device:CPU:0) has insufficient rank for batching.

pred = {"Team1":10764 ,
"Team1Rank":51,
"Team1Prediction":17,
"Team2":8211,
"Team2Rank":81,
"Team2Prediction":1,
"Odds1":1.3,
"Odds2":3.2,
"BestOf":3
}

def pred_fn(features, batch_size=960):
    db = tf.data.Dataset.from_tensor_slices((dict(features))).batch(batch_size)
    return db

predictions = classifier.predict(
    input_fn=lambda: pred_fn(pred))

print(list(predictions))

我遵循标准指南，尝试不同的事情，但没有任何效果

Answer 1

以下更改修复了该问题

db = tf.data.Dataset.from_tensor_slices((dict(features))).batch(batch_size)

到

db = tf.data.Dataset.from_tensors(dict(features)).batch(batch_size)