Chapter 11, Part 1: TextVectorization with output_mode="tf_idf"

[liganega]

The 24th code cell raises the following error when it runs on a PC using Anaconda(Python 3.8.5 + Tensorflow 2.6 or 2.7) while it runs well on Google Colab.

---------------------------------------------------------------------------
InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-25-6747a8415a37> in <module>
----> 1 text_vectorization.adapt(text_only_train_ds)
      2 
      3 tfidf_2gram_train_ds = train_ds.map(lambda x, y: (text_vectorization(x), y))
      4 tfidf_2gram_val_ds = val_ds.map(lambda x, y: (text_vectorization(x), y))
      5 tfidf_2gram_test_ds = test_ds.map(lambda x, y: (text_vectorization(x), y))

~\anaconda3\lib\site-packages\keras\engine\base_preprocessing_layer.py in adapt(self, data, batch_size, steps)
    242       with data_handler.catch_stop_iteration():
    243         for _ in data_handler.steps():
--> 244           self._adapt_function(iterator)
    245           if data_handler.should_sync:
    246             context.async_wait()
...

When the output_mode is not "tf_idf", then everything is ok. The error occurred first with TF 2.6, but it continues to happen even with TF 2.7.

text_vectorization = TextVectorization(
    ngrams=2,
    max_tokens=20000,
    output_mode="tf_idf",
)

The python environment is as follows:

OS: Windows 11 (no WSL2)
Anaconda + TF 2.6 or 2.7

I am wondering WHY!

Many thanks in advance for any help.

[liganega]

I was wrong with Google Colab. The same error occurs even on Google Colab.
(I don't know why it worked yesterday or maybe I believed so at least.)

[liganega]

I think I found the main reason for the confusion.

Whether the error occurs depends on the GPU support.

• with GPU support: Yes, the error occurs.
• without GPU support: No error.

[jmbo1190]

Hello, I've observed the same error on GPU only: in cell with code from Listing 11.11 (training new model with TF-IDF bigram model)

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): No, using code from Deep Learning with Python, 2nd edition, Listing 11.11 Training and testing the TF-IDF bigram model
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 20.04, release='5.4.0-90-generic', version='#101-Ubuntu SMP Fri Oct 15 20:00:55 UTC 2021', machine='x86_64', processor='x86_64'
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device: NA
• TensorFlow installed from (source or binary): binary:
  https://conda.anaconda.org/conda-forge/linux-64/tensorflow-base-2.6.2-cuda112py37h8d33417_2.tar.bz2
  https://conda.anaconda.org/conda-forge/linux-64/tensorflow-estimator-2.6.2-cuda112py37h474db6c_2.tar.bz2
  https://conda.anaconda.org/conda-forge/linux-64/tensorflow-2.6.2-cuda112py37h474db6c_2.tar.bz2
  https://conda.anaconda.org/conda-forge/linux-64/tensorflow-gpu-2.6.2-cuda112py37h0bbbad9_2.tar.bz2
• TensorFlow version (use command below): unknown 2.6.2
• Python version: Python 3.7.12
• Bazel version (if compiling from source):
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version: 11.2/8201
• GPU model and memory: GeForce GTX 1660 SUPER, 5944MiB

You can collect some of this information using our environment capture
script
You can also obtain the TensorFlow version with:

1. TF 1.0: python -c "import tensorflow as tf; print(tf.GIT_VERSION, tf.VERSION)"
2. TF 2.0: python -c "import tensorflow as tf; print(tf.version.GIT_VERSION, tf.version.VERSION)"

Describe the current behavior Error message:

/tmp/ipykernel_901495/519244469.py in <module>
----> 1 text_vectorization.adapt(text_only_train_ds)
      2 
      3 tfidf_2gram_train_ds = train_ds.map(
      4     lambda x, y: (text_vectorization(x), y),
      5     num_parallel_calls=4)

see full log below

Describe the expected behavior
The code should run without errors.

Contributing

• Do you want to contribute a PR? (yes/no): no
• Briefly describe your candidate solution(if contributing):

Standalone code to reproduce the issue
Provide a reproducible test case that is the bare minimum necessary to generate
the problem. If possible, please share a link to Colab/Jupyter/any notebook.
Original code from repo

Other info / logs Include any logs or source code that would be helpful to
diagnose the problem. If including tracebacks, please include the full
traceback. Large logs and files should be attached.

• GPU run log: chapter11_part01_introduction_reprex.md
• CPU run produces no error

[jasonbrancazio]

I just ran into similar issues running on Google Colab with a GPU.

Stacktrace:

/usr/local/lib/python3.7/dist-packages/keras/engine/base_preprocessing_layer.py in adapt(self, data, batch_size, steps)
    242       with data_handler.catch_stop_iteration():
    243         for _ in data_handler.steps():
--> 244           self._adapt_function(iterator)
    245           if data_handler.should_sync:
    246             context.async_wait()

/usr/local/lib/python3.7/dist-packages/tensorflow/python/util/traceback_utils.py in error_handler(*args, **kwargs)
    151     except Exception as e:
    152       filtered_tb = _process_traceback_frames(e.__traceback__)
--> 153       raise e.with_traceback(filtered_tb) from None
    154     finally:
    155       del filtered_tb

/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     57     ctx.ensure_initialized()
     58     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
---> 59                                         inputs, attrs, num_outputs)
     60   except core._NotOkStatusException as e:
     61     if name is not None:

InvalidArgumentError: 2 root error(s) found.
  (0) INVALID_ARGUMENT:  During Variant Host->Device Copy: non-DMA-copy attempted of tensor type: string
	 [[{{node map/TensorArrayUnstack/TensorListFromTensor/_42}}]]
	 [[Func/map/while/body/_1/input/_50/_58]]
  (1) INVALID_ARGUMENT:  During Variant Host->Device Copy: non-DMA-copy attempted of tensor type: string
	 [[{{node map/TensorArrayUnstack/TensorListFromTensor/_42}}]]
0 successful operations.
0 derived errors ignored. [Op:__inference_adapt_step_77044]

Function call stack:
adapt_step -> adapt_step

[SteffenBauer]

I just stumbled over the same problem when trying to run the code in a Jupyter lab notebook on a Jetson Nano (TF 2.7)

I managed to get it to work there by specifying CPU as the device for the adapt operation:

with tf.device("cpu"):
    text_vectorization.adapt(text_only_train_ds)

[ifond]

YES!!! I have received your E-mail——Steven Lee