Hands-On Machine Learning with TensorFlow.js

Converting a TensorFlow SavedModel

Once the SavedModel is created, you can convert the SavedModel into TensorFlow.js format as follows:

$ tensorflowjs_converter \   --output_node_names=output \   --input_format=tf_saved_model \  ./my_saved_model ./my_tfjs_model

The input path and output path are required as positional arguments. (my_saved_model and my_tfjs_model). The model will be generated in the my_tfjs_model directory. The options specified in the preceding example are the minimum requirements. The input_format option is especially important to specify the format explicitly. tfjs-converter automatically infers the compatible output format from this value.

The following is a table showing all the options that can be used with tfjs-converter ...