在 tensorflow 2.4 中使用 sampled_softmax 时无法将符号 Keras 输入/输出转换为 numpy 数组 TypeError

Posted 2023-03-12

【中文标题】在 tensorflow 2.4 中使用 sampled_softmax 时无法将符号 Keras 输入/输出转换为 numpy 数组 TypeError

【英文标题】：Cannot convert a symbolic Keras input/output to a numpy array TypeError when using sampled_softmax in tensorflow 2.4

【发布时间】：2021-03-29 15:54:22

【问题描述】：

我正在尝试使用带有 Keras 的 TF2.4 并使用 tf.nn.sampled_softmax_loss 来训练一个词嵌入分类器。但是，在调用模型的 fit 方法时，会出现“无法将符号 Keras 输入/输出转换为 numpy 数组”的 TypeError。请帮助我修复错误或使用替代方法进行候选抽样。

import tensorflow as tf
import numpy as np

 
TextVectorization = tf.keras.layers.experimental.preprocessing.TextVectorization

class SampledSoftmaxLoss: #(tf.keras.losses.Loss):

  def __init__(self, model, n_classes):
    self.model = model
    output_layer = model.layers[-1]
    self.input = output_layer.input
    self.weights = output_layer.weights
    self.n_classes = n_classes


  def loss(self, y_true, y_pred, **kwargs):
    labels = tf.argmax(y_true, axis=1)
    labels = tf.expand_dims(labels, -1)
    loss = tf.nn.sampled_softmax_loss(
        weights=self.weights[0],
        biases=self.weights[1],
        labels=labels,
        inputs=self.input,
        num_sampled = 3,
        num_classes = self.n_classes
    )
    return loss


max_features = 50  # Maximum vocab size.
max_len = 10  # Sequence length to pad the outputs to.
embedding_dims = 5

input_data = np.array([ 
    "Python Machine Learning",
    "Data Science from Scratch: First Principles with Python", 
    "Hands-On Machine Learning with Scikit-Learn and TensorFlow: Concepts, Tools, and Techniques for Building Intelligent Systems",
    "Introduction to Machine Learning with Python: A Guide for Data Scientists",
    "Vital Introduction to Machine Learning with Python: Best Practices to Improve and Optimize Machine Learning Systems and Algorithms",
    "Machine Learning in Python: Essential Techniques for Predictive Analysis",
    "Python Data Science Handbook: Essential Tools for Working with Data",
    "Introducing Data Science: Big Data, Machine Learning, and more, using Python tools",
    "Real-World Machine Learning"])

labels_one_hot = []
for i in range(len(input_data)):
    labels = np.random.randint(0, 6, max_features)
    labels[labels>1] = 1
    labels_one_hot.append(labels)

labels_one_hot = np.array(labels_one_hot)
 
# Create the text categorical encoding layer.
vectorize_layer = TextVectorization(
         max_tokens=max_features,
         output_mode='int',
         output_sequence_length=max_len)
 
vectorize_layer.adapt(text_dataset)
inp = tf.keras.Input(shape=(1,), dtype=tf.string)
idxs = vectorize_layer(inp)
embed = tf.keras.layers.Embedding(max_features + 1, embedding_dims,input_length=max_len)(idxs)
flat = tf.keras.layers.Flatten()(embed)
out = tf.keras.layers.Dense(labels_one_hot.shape[1])(flat)

model = tf.keras.models.Model(inp, out)

softmax = SampledSoftmaxLoss(model, labels_one_hot.shape[1])

model.compile('adam', loss=softmax.loss)
model.summary()
model.fit(input_data, labels_one_hot)
model.predict(input_data)

这是日志：

INFO:tensorflow:Assets written to: model/assets
Model: "model_5"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_84 (InputLayer)        [(None, 1)]               0         
_________________________________________________________________
text_vectorization_83 (TextV (None, 10)                0         
_________________________________________________________________
embedding_83 (Embedding)     (None, 10, 5)             255       
_________________________________________________________________
flatten_25 (Flatten)         (None, 50)                0         
_________________________________________________________________
dense_69 (Dense)             (None, 50)                2550      
=================================================================
Total params: 2,805
Trainable params: 2,805
Non-trainable params: 0
_________________________________________________________________

---------------------------------------------------------------------------

TypeError                                 Traceback (most recent call last)

<ipython-input-87-c35b8d1321a1> in <module>()
    106 model.compile('adam', loss=softmax.loss)
    107 model.summary()
--> 108 model.fit(input_data, labels_one_hot)
    109 # Now, the model can map strings to integers, and you can add an embedding
    110 # layer to map these integers to learned embeddings.

9 frames

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py in wrapper(*args, **kwargs)
    975           except Exception as e:  # pylint:disable=broad-except
    976             if hasattr(e, "ag_error_metadata"):
--> 977               raise e.ag_error_metadata.to_exception(e)
    978             else:
    979               raise

TypeError: in user code:

    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:805 train_function  *
        return step_function(self, iterator)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:795 step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:1259 run
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2730 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:3417 _call_for_each_replica
        return fn(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:788 run_step  **
        outputs = model.train_step(data)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:756 train_step
        y, y_pred, sample_weight, regularization_losses=self.losses)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/compile_utils.py:238 __call__
        total_loss_metric_value, sample_weight=batch_dim)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/utils/metrics_utils.py:90 decorated
        update_op = update_state_fn(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/metrics.py:177 update_state_fn
        return ag_update_state(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/metrics.py:364 update_state  **
        sample_weight, values)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/weights_broadcast_ops.py:155 broadcast_weights
        values = ops.convert_to_tensor(values, name="values")
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/profiler/trace.py:163 wrapped
        return func(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py:1540 convert_to_tensor
        ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py:339 _constant_tensor_conversion_function
        return constant(v, dtype=dtype, name=name)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py:265 constant
        allow_broadcast=True)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py:283 _constant_impl
        allow_broadcast=allow_broadcast))
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_util.py:435 make_tensor_proto
        values = np.asarray(values)
    /usr/local/lib/python3.6/dist-packages/numpy/core/_asarray.py:83 asarray
        return array(a, dtype, copy=False, order=order)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/keras_tensor.py:274 __array__
        'Cannot convert a symbolic Keras input/output to a numpy array. '

    TypeError: Cannot convert a symbolic Keras input/output to a numpy array. This error may indicate that you're trying to pass a symbolic value to a NumPy call, which is not supported. Or, you may be trying to pass Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing Keras from automatically converting the API call to a lambda layer in the Functional Model.

【问题讨论】：

你检查this question了吗？可能与您的问题有关

这可能是由于混合了 numpy 和 Tensorflow ops 数据类型等输入。您可以按照此链接和提到的教程检查TextVectorization 在不同解决方案中的用法。 tensorflow.org/api_docs/python/tf/keras/layers/experimental/…

【参考方案1】：

出现此错误的主要原因之一是您可能在 TensorFlow 之前安装了 Keras。 如果是这种情况，重新启动内核就可以了

【讨论】：

根本不行，我只安装了Tensorflow，用的是tensorflow包里的kreas

【参考方案2】：

错误是由您的自定义损失函数引起的。您应该禁用 TF 急切执行模式。

from tensorflow.python.framework.ops import disable_eager_execution

disable_eager_execution()

另外，在model.compile(...) 中使用experimental_run_tf_function=False

【讨论】：

这不是解决方案，在我的情况下，我在 GPU 上运行带有 LSTM 层的模型，一旦我禁用了急切执行，另一个错误是 LSTM 无法使用 GPU，它不遵守标准

experimental_run_tf_function=False in model.compile(...) 可以在没有 disable_eager_execution() 的情况下提供帮助。它对你有什么影响吗？

我遇到了同样的问题。 disable_eager_execution() 有效，但使用后出现前置条件错误。

【参考方案3】：

以下解决方案让我有点头疼，但我不确定它为什么有效。 我是 DL 新手，我只是想提供帮助。

试试：

del model

之前：

model = tf.keras.models.Model(inp, out)

我知道我可能错了，欢迎教育我，我愿意学习:)。