is it tensorflow example bug ?

[Q82822]

System information

Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
OS Platform and Distribution (e.g., Linux Ubuntu 16.04): window10 and jupyter
Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device: iPhone X
TensorFlow installed from (source or binary): source
TensorFlow version (use command below): 1.11
Python version: 3.6
Bazel version (if compiling from source):
GCC/Compiler version (if compiling from source):
CUDA/cuDNN version: 9.1 / 7.1
GPU model and memory:
Exact command to reproduce:

Describe the problem

I used the tensorflow github example form here change the CNN model
img = tf.keras.applications.inception_v3.preprocess_input(img) --->img = tf.keras.applications.inception_resnet_v2.preprocess_input(img)

image_model = tf.keras.applications.InceptionV3(include_top=False, weights='imagenet') ---->image_model =tf.keras.applications.InceptionResNetV2(include_top=False, weights='imagenet')

the output

Epoch 1 Batch 0 Loss nan
Epoch 1 Batch 100 Loss nan
Epoch 1 Batch 200 Loss nan
Epoch 1 Batch 300 Loss nan
Epoch 1 Loss nan

i have tried to change the learning rate and loss function,even others CNN model (vgg19,vgg16) but it was same

Source code / logs

the Source code is tensorflow example from here
you just need to change the CNN model and wait a few minute run it.

[Harshini-Gadige]

@drpngx Hi, could you please look into this issue.

[drpngx]

@Harshini-Gadige did you try without modifications?

[Harshini-Gadige]

@drpngx Yes I tried the original one and it took long time to load the data and then the system crashed.

[drpngx]

@MarkDaoust Is that a known documentation problem?

[Q82822]

@drpngx So this example only can use InceptionV3 Model? have any solution?

thanks!

[drpngx]

I think it's probably that the docs need to be refreshed. I'm not sure what's the plan for this.

[Harshini-Gadige]

@MarkDaoust Hi, could you please look into this ?

[Q82822]

hi, this problem was figure out in new tensorflow 1.12 , but it stall have another problem.
In this code will print error
"provided list of inputs contains objects other than 'EagerTensor'. Item 0 is Tensor"
add it happen in use Vgg16 model.

This is the code here
change the CNN model

img = tf.keras.applications.inception_v3.preprocess_input(img) --->img = tf.keras.applications.vgg16.preprocess_input(img)

image_model = tf.keras.applications.InceptionV3(include_top=False, weights='imagenet') ---->image_model =tf.keras.applications.Vgg16(include_top=False, weights='imagenet')

 img = tf.image.resize_images(img, (299, 299))---> img = tf.image.resize_images(img, (224, 224))

change the image features_shape & attention_features_shape

features_shape = 2048 -->512
attention_features_shape = 64-->81

problem code in here

# captions on the validation set
rid = np.random.randint(0, len(img_name_val))
image = img_name_val[rid]
real_caption = ' '.join([index_word[i] for i in cap_val[rid] if i not in [0]])
result, attention_plot = evaluate(image)

print ('Real Caption:', real_caption)
print ('Prediction Caption:', ' '.join(result))
plot_attention(image, result, attention_plot)
# opening the image
`Image.open(img_name_val[rid])`

and error here

FallbackException                        Traceback (most recent call last)
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_array_ops.py in reshape(tensor, shape, name)
   6494         _ctx._context_handle, _ctx._eager_context.device_name, "Reshape",
-> 6495         name, _ctx._post_execution_callbacks, tensor, shape)
   6496       return _result

_FallbackException: This function does not handle the case of the path where all inputs are not already EagerTensors.

During handling of the above exception, another exception occurred:

TypeError                                 Traceback (most recent call last)
<ipython-input-31-98d70c233c4f> in <module>()
      2 image = img_name_val[rid]
      3 real_caption = ' '.join([index_word[i] for i in cap_val[rid] if i not in [0]])
----> 4 result, attention_plot = evaluate(image)
      5 
      6 print ('Real Caption:', real_caption)

<ipython-input-29-55980383ccd6> in evaluate(image)
      4     hidden = decoder.reset_state(batch_size=1)
      5 
----> 6     temp_input = tf.expand_dims(load_image(image)[0], 0)
      7     img_tensor_val = image_features_extract_model(temp_input)
      8     img_tensor_val = tf.reshape(img_tensor_val, (img_tensor_val.shape[0], -1, img_tensor_val.shape[3]))

<ipython-input-5-28f690dfd033> in load_image(image_path)
      3     img = tf.image.decode_jpeg(img, channels=3)
      4     img = tf.image.resize_images(img, (224, 224))
----> 5     img = tf.keras.applications.vgg16.preprocess_input(img)
      6     return img, image_path

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/applications/__init__.py in wrapper(*args, **kwargs)
     68       kwargs['models'] = models
     69       kwargs['utils'] = utils
---> 70     return base_fun(*args, **kwargs)
     71   return wrapper
     72 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/applications/vgg16.py in preprocess_input(*args, **kwargs)
     42 @keras_modules_injection
     43 def preprocess_input(*args, **kwargs):
---> 44   return vgg16.preprocess_input(*args, **kwargs)

/usr/local/lib/python3.6/dist-packages/keras_applications/imagenet_utils.py in preprocess_input(x, data_format, mode, **kwargs)
    193     else:
    194         return _preprocess_symbolic_input(x, data_format=data_format,
--> 195                                           mode=mode, **kwargs)
    196 
    197 

/usr/local/lib/python3.6/dist-packages/keras_applications/imagenet_utils.py in _preprocess_symbolic_input(x, data_format, mode, **kwargs)
    149             data_format=data_format)
    150     else:
--> 151         x = backend.bias_add(x, _IMAGENET_MEAN, data_format)
    152     if std is not None:
    153         x /= std

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/backend.py in bias_add(x, bias, data_format)
   4595     elif data_format == 'channels_last':
   4596       if len(bias_shape) == 1:
-> 4597         x = x + reshape(bias, (1, 1, bias_shape[0]))
   4598       else:
   4599         x = x + reshape(bias, (1,) + bias_shape)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/backend.py in reshape(x, shape)
   2279       A tensor.
   2280   """
-> 2281   return array_ops.reshape(x, shape)
   2282 
   2283 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_array_ops.py in reshape(tensor, shape, name)
   6497     except _core._FallbackException:
   6498       return reshape_eager_fallback(
-> 6499           tensor, shape, name=name, ctx=_ctx)
   6500     except _core._NotOkStatusException as e:
   6501       if name is not None:

/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_array_ops.py in reshape_eager_fallback(tensor, shape, name, ctx)
   6516   _attrs = ("T", _attr_T, "Tshape", _attr_Tshape)
   6517   _result = _execute.execute(b"Reshape", 1, inputs=_inputs_flat, attrs=_attrs,
-> 6518                              ctx=_ctx, name=name)
   6519   _execute.record_gradient(
   6520       "Reshape", _inputs_flat, _attrs, _result, name)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     58     tensors = pywrap_tensorflow.TFE_Py_Execute(ctx._handle, device_name,
     59                                                op_name, inputs, attrs,
---> 60                                                num_outputs)
     61   except core._NotOkStatusException as e:
     62     if name is not None:

TypeError: provided list of inputs contains objects other than 'EagerTensor'. Item 0 is Tensor

THX!!

[yashk2810]

Can you try this out: https://www.tensorflow.org/tutorials/text/image_captioning

I am closing this. Please reopen if still failing.