[tfjs-converter] Support for ops: `BitwiseXor`, `Bitcast`, `LeftShift`, `BitwiseOr`, `RightShift`

[josephrocca]

I'm converting a BART-based model from TensorFlow Saved Model format, and the lack of support for these ops is preventing the convesion: BitwiseXor, Bitcast, LeftShift, BitwiseOr, RightShift.

System information

• TensorFlow.js version (you are using): tensorflowjs==3.18.0 (python converter package)
• Are you willing to contribute it (Yes/No): Yes, willing to try to help if someone can provide guidance

Alternatives Another possible route to solving this issue is adding "TF Select" ops (tf.lite.OpsSet.SELECT_TF_OPS) to the tfjs-tflite runtime: https://github.com/tensorflow/tfjs/issues/5844 The missing ops from the tfjs-tflite runtime are Bitcast, BitwiseOr, BitwiseXor, ConcatV2, Erf, LeftShift, RightShift, ScatterNd, SelectV2, StridedSlice, but these are all TF Select ops, so the model would be able to be used in the browser if the tfjs-tflite runtime supported TF Select ops.

[pyu10055]

@josephrocca Thank you for submitting this issue and willing to contribute to the library. Here is the doc for adding an new TF op to TFJS Please let us know if you have any questions in following the process. Thanks

[gaikwadrahul8]

Hi, @josephrocca

Apologize for the delayed response and we are re-visiting our older feature requests and checking whether those feature requests implemented or not as of now and you shown interest to contribute for these missing Ops in TFJs so May I know are you working on this Ops in TFJs please ?

If someone wants to contribute for this feature then you're always welcome and please feel free to do and please refer these links Ref-1, Ref-2 . Thank you!

[josephrocca]

Hi, yes these are still needed, so please keep this open, but I am unable to work on them.