Sep 10, 2020 · During the conversion process we traverse the model graph and check that each operation is supported by TensorFlow.js. If so, we write the graph into a format that the browser can consume. We try to optimize the model for being served on the web by sharding the weights into 4MB files - that way they can be cached by browsers. Nov 24, 2020 · ONNX Model to TensorFlow SavedModel. Refer to this notebook for complete code details mentioned in this section. As mentioned earlier, the best feature of ONNX is interoperability. Once we have the access to the ONNX model we can convert it into any other existing popular frameworks very easily.
Onnx-Tensorflow doesn't seem to be setting tags on the pb model. I haven't been able to find a flag in the API that sets this. This is making it tricky to view the model in Tensorboard.