Transcription generation(last cell) in the TFlite conversion notebook throws error

[zhubarb]

I tried running the tflite conversion notebook under /models in colab. It all runs OK until the final cell:

# loaded model... now with generate!
tflite_model_path = 'whisper-tiny.en.tflite'
interpreter = tf.lite.Interpreter(tflite_model_path)

tflite_generate = interpreter.get_signature_runner()
generated_ids = tflite_generate(input_features=input_features)["sequences"]
transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

Here, I get the error. To reproduce, you can run the notebook front to back in colab, and you should see the error. Any ideas why this is happening? Is it a version issue. What were the original package versions for !pip install transformers !pip install datasets?

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

RuntimeError                              Traceback (most recent call last)

[<ipython-input-5-aff8cd54abe2>](https://localhost:8080/#) in <cell line: 6>()
      4 
      5 tflite_generate = interpreter.get_signature_runner()
----> 6 generated_ids = tflite_generate(input_features=input_features)["sequences"]
      7 transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
      8 transcription

[/usr/local/lib/python3.10/dist-packages/tensorflow/lite/python/interpreter.py](https://localhost:8080/#) in __call__(self, **kwargs)
    247                                           self._subgraph_index)
    248 
--> 249     self._interpreter_wrapper.Invoke(self._subgraph_index)
    250     result = {}
    251     for output_name, output_index in self._outputs:

RuntimeError: tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::apply(optimized_ops::Mean<T, U>, args) was not true.tensorflow/lite/kernels/reduce.cc:445 std::appl...

[zhubarb]

It is a versioning error as expected. Tensorflow 15.0 messes everything up. Not the first time it is getting in my way(!).

Replace the import with

!pip install transformers==4.33.0
!pip install datasets
!pip install tensorflow==2.14.0

And the notebook runs without errors.