Interactive cli for python converter

README.md

@@ -25,12 +25,10 @@ __0. Please make sure that you run in a Docker container or a virtual environmen
  The script pulls its own subset of TensorFlow, which might conflict with the
  existing TensorFlow/Keras installation.
 
-__Note__: *Check that [`tf-nightly-2.0-preview`](https://pypi.org/project/tf-nightly-2.0-preview/#files) is available for your platform.*
-
-Most of the times, this means that you have to use Python 3.6.8 in your local
-environment. To force Python 3.6.8 in your local project, you can install
-[`pyenv`](https://github.com/pyenv/pyenv) and proceed as follows in the target
-directory:
+The converter supports both Python 2 and Python 3. But Python 3.6.8 is recommended
+for your local environment. To force Python 3.6.8 in your local project, you can
+install [`pyenv`](https://github.com/pyenv/pyenv) and proceed as follows in the
+target directory:
 
 ```bash
 pyenv install 3.6.8
@@ -52,7 +50,35 @@ __1. Install the TensorFlow.js pip package:__
  pip install tensorflowjs
 ```
 
-__2. Run the converter script provided by the pip package:__
+__2. Run the conversion script provided by the pip package:__
+
+There are two way to trigger the model conversion:
+
+- The conversion wizard: tensorflowjs_wizard
+- Regular conversion script: tensorflowjs_converter
+
+To start the conversion wizard:
+```bash
+tensorflowjs_wizard
+```
+
+This tool will walk you through the conversion process and provide you with
+details explanations for each choice you need to make. Behind the scene it calls
+the converter script (`tensorflowjs_converter`) in pip package. This is the easier
+way to convert a single model.
+
+There is also dry run mode for the wizard, which will not perform the actual
+conversion but only generate the command for `tensorflowjs_converter` command.
+This command can be used in your own shell script.
+
+```bash
+tensorflowjs_wizard --dryrun
+```
+
+To convert a batch of models or integrate the conversion process into your own
+script, you should look into using the tensorflowjs_converter script.
+
+Here is detail information of parameters of the converter script.
 
 The converter expects a __TensorFlow SavedModel__, __TensorFlow Hub module__,
 __TensorFlow.js JSON__ format, __Keras HDF5 model__, or __tf.keras SavedModel__

python/requirements.txt

@@ -4,3 +4,4 @@ numpy==1.16.4
 six==1.11.0
 tensorflow==1.14.0
 tensorflow-hub==0.5.0
+PyInquirer==1.0.3

python/setup.py

@@ -27,6 +27,7 @@ def _get_requirements(file):
 
 CONSOLE_SCRIPTS = [
     'tensorflowjs_converter = tensorflowjs.converters.converter:pip_main',
+    'tensorflowjs_wizard = tensorflowjs.wizard:main',
 ]
 
 setuptools.setup(
@@ -57,6 +58,7 @@ def _get_requirements(file):
         'tensorflowjs.version',
         'tensorflowjs.quantization',
         'tensorflowjs.read_weights',
+        'tensorflowjs.wizard',
         'tensorflowjs.write_weights',
         'tensorflowjs.converters',
         'tensorflowjs.converters.common',

python/tensorflowjs/BUILD

@@ -48,6 +48,13 @@ py_library(
     # `pip install tensorflow` or `pip install tensorflow-gpu`.
 )
 
+py_library(
+    name = "expect_PyInquirer_installed",
+    # This is a dummy rule used as a PyInquirer dependency in open-source.
+    # We expect PyInquirer to already be installed on the system, e.g. via
+    # `pip install PyInquirer`.
+)
+
 py_library(
     name = "quantization",
     srcs = ["quantization.py"],
@@ -116,6 +123,29 @@ py_test(
     ],
 )
 
+py_test(
+    name = "wizard_test",
+    srcs = ["wizard_test.py"],
+    srcs_version = "PY2AND3",
+    deps = [
+        ":expect_numpy_installed",
+        ":wizard",
+    ],
+)
+
+py_binary(
+    name = "wizard",
+    srcs = ["wizard.py"],
+    srcs_version = "PY2AND3",
+    deps = [
+        ":converters/common",
+        ":converters/converter",
+        "//tensorflowjs:expect_h5py_installed",
+        "//tensorflowjs:expect_keras_installed",
+        "//tensorflowjs:expect_tensorflow_installed",
+    ],
+)
+
 # A filegroup BUILD target that includes all the op list json files in the
 # the op_list/ folder. The op_list folder itself is a symbolic link to the
 # actual op_list folder under src/.

python/tensorflowjs/__init__.py

@@ -21,5 +21,6 @@
 from tensorflowjs import converters
 from tensorflowjs import quantization
 from tensorflowjs import version
+from tensorflowjs import wizard
 
 __version__ = version.version

python/tensorflowjs/converters/BUILD

@@ -91,6 +91,7 @@ py_binary(
     srcs = ["converter.py"],
     srcs_version = "PY2AND3",
     deps = [
+        ":common",
         ":keras_h5_conversion",
         ":keras_tfjs_loader",
         ":tf_saved_model_conversion_v2",

python/tensorflowjs/converters/common.py

@@ -30,6 +30,27 @@
 GENERATED_BY_KEY = 'generatedBy'
 CONVERTED_BY_KEY = 'convertedBy'
 
+# Model formats.
+KERAS_SAVED_MODEL = 'keras_saved_model'
+KERAS_MODEL = 'keras'
+TF_SAVED_MODEL = 'tf_saved_model'
+TF_HUB_MODEL = 'tf_hub'
+TFJS_GRAPH_MODEL = 'tfjs_graph_model'
+TFJS_LAYERS_MODEL = 'tfjs_layers_model'
+
+# CLI argument strings.
+INPUT_PATH = 'input_path'
+OUTPUT_PATH = 'output_path'
+INPUT_FORMAT = 'input_format'
+OUTPUT_FORMAT = 'output_format'
+SIGNATURE_NAME = 'signature_name'
+SAVED_MODEL_TAGS = 'saved_model_tags'
+QUANTIZATION_BYTES = 'quantization_bytes'
+SPLIT_WEIGHTS_BY_LAYER = 'split_weights_by_layer'
+VERSION = 'version'
+SKIP_OP_CHECK = 'skip_op_check'
+STRIP_DEBUG_OPS = 'strip_debug_ops'
+WEIGHT_SHARD_SIZE_BYTES = 'weight_shard_size_bytes'
 
 def get_converted_by():
   """Get the convertedBy string for storage in model artifacts."""