Tensorflow 2.3 model.evaluate报错InvalidArgumentError: Incompatible shapes: [1,64] vs. [1,128]

⚡插播一条老家自产的糖心苹果，多个品种，欢迎选购！有问题随时私信我⚡：🍎🍎来自雪域高原的馈赠——海拔2000米的大凉山高原生态糖心苹果，欢迎选购！！🍎🍎

大凉山高原生态糖心苹果

1.软件环境⚙️

Windows10 教育版64位
Python 3.6.3
Tensorflow-GPU 2.3.0
CUDA 10.1

2.问题描述🔍

我们在模型训练完时，都需要对模型的性能进行评估。而在Tensorflow.Keras中，往往通过.flow_from_directory函数读入本地的图片，然后使用model.evaluate对模型进行精度评估：

test_datagen = ImageDataGenerator(preprocessing_function=preprocessing_function)

test_generator = test_datagen.flow_from_directory(test_dir,
                                                  shuffle=False,
                                                  target_size=(299,299),
                                                  batch_size=32)

print("================开始模型评估======================")
model_evaluation = model.evaluate(test_generator, verbose=1)

比如我们这边有一个评估数据集val-fewer-sample，该数据集中包含dog、cat两类：

这些样本已经被打上了正确的标签（即文件夹名），我们训练出来的分类器对这些样本进行预测，如果标签对得上，那么该图片预测正确。
如果你训练的时候使用的是softmax那么不会有问题，但二分类问题我更喜欢用sigmoid，这个时候，如果是Tensorflow 2.3，就会出现报错：

tensorflow.python.framework.errors_impl.InvalidArgumentError: Incompatible shapes: [1,64] vs. [1,128]

即：

Traceback (most recent call last):
  File "<input>", line 1, in <module>
  File "C:\Program Files\JetBrains\PyCharm 2020.1\plugins\python\helpers\pydev\_pydev_bundle\pydev_umd.py", line 197, in runfile
    pydev_imports.execfile(filename, global_vars, local_vars)  # execute the script
  File "C:\Program Files\JetBrains\PyCharm 2020.1\plugins\python\helpers\pydev\_pydev_imps\_pydev_execfile.py", line 18, in execfile
    exec(compile(contents+"\n", file, 'exec'), glob, loc)
  File "E:/Code/Python/classification model evaluation/model_evaluation_sigmoid.py", line 71, in <module>
    model_evaluation = model.evaluate(test_generator, verbose=1, workers=4, return_dict=True)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\keras\engine\training.py", line 108, in _method_wrapper
    return method(self, *args, **kwargs)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1379, in evaluate
    tmp_logs = test_function(iterator)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\def_function.py", line 780, in __call__
    result = self._call(*args, **kwds)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\def_function.py", line 846, in _call
    return self._concrete_stateful_fn._filtered_call(canon_args, canon_kwds)  # pylint: disable=protected-access
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\function.py", line 1848, in _filtered_call
    cancellation_manager=cancellation_manager)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\function.py", line 1924, in _call_flat
    ctx, args, cancellation_manager=cancellation_manager))
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\function.py", line 550, in call
    ctx=ctx)
  File "C:\Users\Jayce\Anaconda3\envs\tf2.3\lib\site-packages\tensorflow\python\eager\execute.py", line 60, in quick_execute
    inputs, attrs, num_outputs)
tensorflow.python.framework.errors_impl.InvalidArgumentError:  Incompatible shapes: [1,64] vs. [1,128]
	 [[node LogicalAnd (defined at E:/Code/Python/classification model evaluation/model_evaluation_sigmoid.py:71) ]] [Op:__inference_test_function_7318]
Function call stack:
test_function

可以看到，2个版本的Tensorflow 的报错原因都是tensor的shape不一致，这就很尴尬了，为什么softmax训练出来的模型可以正常评估，但是sigmoid训练出来的模型就报错呢？

3.解决方法🐡

经过查询，发现是因为.flow_from_directory中的class_mode参数的默认值是categorical，而我们使用的是sigmoid进行训练：

  def flow_from_directory(self,
                          directory,
                          target_size=(256, 256),
                          color_mode='rgb',
                          classes=None,
                          class_mode='categorical',
                          batch_size=32,
                          shuffle=True,
                          seed=None,
                          save_to_dir=None,
                          save_prefix='',
                          save_format='png',
                          follow_links=False,
                          subset=None,
                          interpolation='nearest'):
    """Takes the path to a directory & generates batches of augmented data."""

因此需要将class_mode修改为binary才能用sigmoid适配，即：

test_generator = test_datagen.flow_from_directory(test_dir,
                                                  shuffle=False,
                                                  target_size=(299,299),
                                                  batch_size=32)
# 修改为：
test_generator = test_datagen.flow_from_directory(test_dir,
                                                  shuffle=False,
                                                  target_size=(299,299),
                                                  class_mode='binary',
                                                  batch_size=32)

4.结果预览🤔

修改完class_mode之后，发现模型评估可以正常运行了：

Found 12226 images belonging to 2 classes.
================开始模型评估======================
2022-09-01 13:56:51.913965: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cublas64_10.dll
2022-09-01 13:56:53.702954: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudnn64_7.dll
C:\Users\Anaconda3\envs\tf2.3\lib\site-packages\PIL\Image.py:974: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  "Palette images with Transparency expressed in bytes should be "
2022-09-01 13:56:55.118644: W tensorflow/stream_executor/gpu/redzone_allocator.cc:314] Internal: Invoking GPU asm compilation is supported on Cuda non-Windows platforms only
Relying on driver to perform ptx compilation. 
Modify $PATH to customize ptxas location.
This message will be only logged once.
 53/192 [=======>......................] - ETA: 28s - loss: 0.2310 - accuracy: 0.9593 - precision: 0.9521 - recall: 0.9668

​
渣男！都看到这里了，还不赶紧点赞，评论，收藏走一波？