search

apple / coremltools

Core ML tools contain supporting tools for Core ML model conversion, editing, and validation.
https://coremltools.readme.io
BSD 3-Clause "New" or "Revised" License
4.38k stars 631 forks source link

NotImplementedError: Conversion for TF op 'Bucketize' not implemented. #1146

Open CaptainDP opened 3 years ago

CaptainDP commented 3 years ago

❓Question

Error encountered while converting the TensorFlow model to CoreML model: NotImplementedError: Conversion for TF op 'Bucketize' not implemented.

TensorFlow uses the following code for training:

numeric_column = tf. Feature_column. Numeric_column (col) buket_col = tf. Feature_column. Bucketized_column (numeric_column, boundary_dict [col]) embedding_column = tf.feature_column. Embedding_column (buket_col, K + 1)

How do I skip support for bucketized_column?

System Information

TobyRoseman commented 3 years ago

Looks like we don't have support for converting tf.feature_column.bucketized_column.

I recommend removing that before converting your model. Then doing the bucketizing in your app before using your Core ML model.

I'll leave this issue open until we support this operation.