tensorflow Keras：維度必須相等 ValueError

Question

我正在嘗試在 Keras 中訓練 model，以在出現棋盤時建議最佳下一步行動。 棋盤表示為 64 個整數的列表（0 表示空，1 表示玩家，2 表示敵人）。 output 由字段列表和該字段上的圖形應該移動的方向表示，這意味着我需要兩個輸出層，大小為 64（字段數）和 5（可能的移動方向數，包括兩個前向並且在游戲結束時沒有移動）。 我有一個板列表和一個解決方案列表。 但是，當我嘗試適合 model 時，出現上述錯誤。

確切的錯誤信息是：

Epoch 1/75
Traceback (most recent call last):
  File "C:\Users\lulll\Documents\CodeStuff\tfTesting\main.py", line 75, in <module>
    model.fit(train_fig_starts, train_fig_moves, epochs=75)
  File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\utils\traceback_utils.py", line 70, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "C:\Users\lulll\AppData\Local\Temp\__autograph_generated_filej0zia4d5.py", line 15, in tf__train_function
    retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
ValueError: in user code:

    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1249, in train_function  *
        return step_function(self, iterator)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1233, in step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1222, in run_step  **
        outputs = model.train_step(data)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1024, in train_step
        loss = self.compute_loss(x, y, y_pred, sample_weight)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1082, in compute_loss
        return self.compiled_loss(
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\compile_utils.py", line 265, in __call__
        loss_value = loss_obj(y_t, y_p, sample_weight=sw)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 152, in __call__
        losses = call_fn(y_true, y_pred)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 284, in call  **
        return ag_fn(y_true, y_pred, **self._fn_kwargs)
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 2176, in binary_crossentropy
        backend.binary_crossentropy(y_true, y_pred, from_logits=from_logits),
    File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\backend.py", line 5688, in binary_crossentropy
        bce = target * tf.math.log(output + epsilon())

    ValueError: Dimensions must be equal, but are 2 and 64 for '{{node binary_crossentropy/mul}} = Mul[T=DT_FLOAT](binary_crossentropy/Cast, binary_crossentropy/Log)' with input shapes: [?,2], [?,64].

我完全不知道是什么原因造成的。 我已經搜索了錯誤，但我發現的唯一提及似乎描述了一個完全不同的場景。 因為它可能有幫助，下面是用於創建和擬合 model 的代碼：

inputs = tf.keras.layers.Input(shape=64)
x = tf.keras.layers.Dense(32, activation='relu')(inputs)
out_field = tf.keras.layers.Dense(64, name="field")(x)
out_movement = tf.keras.layers.Dense(5, name="movement")(x)

model = tf.keras.Model(inputs=inputs, outputs=[out_field, out_movement])

model.compile(optimizer='adam',
              loss=tf.keras.losses.BinaryCrossentropy(),
              metrics=['accuracy'])

model.fit(train_fig_starts, train_fig_moves, epochs=75) #train_fig_starts and moves are defined above

編輯 1：這是我正在使用的數據集的示例（整個內容對於字符限制來說太長了）

train_fig_starts = [[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 2, 0, 1, 0, 0, 0, 0, 1, 2, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 2, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 2, 1, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 0, 0, 0, 0, 1], [0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 2, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0], [0, 2, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 2, 1, 2, 2, 2, 0, 0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 2, 2, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 1, 2, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0]]
train_fig_moves = [[0, 0], [0, 0], [0, 0], [0, 0], [15, 2], [15, 2]]

編輯 2：我將其更改為 sparsecategorialcrossentropy 因為這看起來更像我正在尋找的東西。 這是現在的 model 代碼

inputs = tf.keras.layers.Input(shape=64)
x = tf.keras.layers.Dense(64, activation='relu')(inputs)

out_field = tf.keras.layers.Dense(64, activation="relu",  name="field")(x)
out_field = tf.keras.layers.Dense(64, activation="softmax", name="field_softmax")(out_field)

out_movement = tf.keras.layers.Dense(5, activation="relu", name="movement")(x)
out_movement = tf.keras.layers.Dense(5, activation="softmax", name="movement_softmax")(out_movement)

model = tf.keras.Model(inputs=inputs, outputs=[out_field, out_movement])

print(model.summary())
tf.keras.utils.plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)

model.compile(optimizer='adam',
              loss=[tf.keras.losses.SparseCategoricalCrossentropy(),
                    tf.keras.losses.SparseCategoricalCrossentropy()],
              metrics=['accuracy'])

它仍然會拋出錯誤，這次是以下錯誤：

Node: 'sparse_categorical_crossentropy_1/SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits'
logits and labels must have the same first dimension, got logits shape [32,5] and labels shape [64]
     [[{{node sparse_categorical_crossentropy_1/SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits}}]] [Op:__inference_train_function_1666]

我不知道為什么會這樣。 Output logits 和 labels 都應該是 [64, 2]。 因為我使用的是稀疏交叉熵，所以我應該能夠在我的訓練數據中使用整數來表示具有最高 logit 的輸出神經元的“索引”，對吧？ 如我錯了請糾正我。 如果有幫助，這是我的 model 的圖表： model 的 plot

Answer 1

問題是您從 model 返回 64 位長度的 output，稱為logit ，而您的損失 function 需要 2 位長度的輸入。

import numpy as np
import tensorflow as tf

train_fig_starts = tf.constant([[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 2, 0, 1, 0, 0, 0, 0, 1, 2, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 2, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 2, 1, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 0, 0, 0, 0, 1], [0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 2, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0], [0, 2, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 2, 1, 2, 2, 2, 0, 0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 2, 2, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 1, 2, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0]])
train_fig_moves = tf.constant([[0, 0], [0, 0], [0, 0], [0, 0], [15, 2], [15, 2]])

inputs = tf.keras.layers.Input(shape=64)
x = tf.keras.layers.Dense(32, activation='relu')(inputs)
out_field = tf.keras.layers.Dense(64, activation='relu')(x)
out_field = tf.keras.layers.Dense(2, name="field" , activation='softmax')(out_field)

out_movement = tf.keras.layers.Dense(32, activation='relu')(x)
out_movement = tf.keras.layers.Dense(2, name="movement", activation='softmax')(out_movement)

model = tf.keras.Model(inputs=inputs, outputs=[out_field, out_movement])

model.compile(optimizer='adam',
              loss=tf.keras.losses.CategoricalCrossentropy(),
              metrics=['accuracy'])
model.fit(train_fig_starts, train_fig_moves, epochs=75) #train_fig_starts and moves are defined above

0s 8ms/step - loss: 5.2342 - field_loss: 2.7274 - movement_loss: 2.5068 - field_accuracy: 0.6667 - movement_accuracy: 1.0000

Answer 2

所以我現在自己解決了這個問題。 老實說，這是一個非常愚蠢的錯誤，但錯誤消息並沒有很好地解釋發生了什么。 我將輸出交換為一種熱編碼，並將損失更改為 CategorialCrossEntropy，這也更適合分類問題（由於某些原因，Sparse 不適用於我的整數）。 之后，我需要將 label 列表從包含 len = 2 列表的 1dim 列表更改為包含字段和 move one hots 的 2dim 列表，在單獨的列表中。 如果有人遇到類似的問題並且無法理解它，也許這會有所幫助。