Browser-based TensorFlow.js C1_W3 tensorflowjs_converter

Having a problem with Google Colab

The ungraded lab won’t run the final step to convert the model.

!tensorflowjs_converter --input_format=keras {saved_model_path} ./

It results in: “command not found: tensorflowjs_converter”

I had thought something was incorrect with my local installation but it also doesn’t look to run in Colab.

The final assignment also requires this to run as the last step so it can’t be completed.

Hi Alan! Sorry for the late reply. You might have forgotten to uncomment the first code cell at the top of the notebook. You need to install the tensorflowjs_converter package to make use of the command. Kindly run this first:

!pip install tensorflowjs

Hope this helps!

I have an issue with this assignment; I use this code to save the model:

tf.Keras.experimental.export_saved_model(model, saved_model_path)
and get this error

module ‘keras.api._v2.keras.experimental’ has no attribute ‘export_saved_model’

So use this
and create an h5 model and then use the above command and create JSON and bin’s file, but can’t pass the assignment.

1 Like

I have the same problem:
I completed all the tasks, created the model, converted it to JSON, and uploaded all necessary files to Coucera.
The recommended version of TensorFlow 2.2 is not able to be installed in Colab by the provided code. Maybe here is the reason.
I need help understanding what I did wrong if I followed all the instructions.
The grading system needs to explain the reason.

All I see is: The model in your submission is invalid. Please try again!

Did you solve this? I am on this same challenge right now


Unfortunately I had to cancel my certification. Given that the examples provided seemed to be broken and I was paying a monthly fee in order to get the certification there wasn’t much choice at the time.

Sorry you had to cancel your certification. If you have to run an older version of tensorflow like 2.2, Kaggle supports older versions and the computer engine is good. I did this assignment on Kaggle

I also think the course should be reviewed. There are better ways of using tensorflow now

Same issue. I used 5 different codes in converting the saved Keras model into JSON and still couldn’t pass. Invalid submission

Planning to cancel as well if I can’t pass this too