Open dhananjaisharma10 opened 5 years ago
You could try "logits": https://github.com/tensorflow/tpu/blob/master/models/official/mnasnet/post_quantization.py It shares code with mnasnet. BTW, you can always train the model and export a savedmodel to inspect signatures we defined.
Thanks for the same!
I'm trying to convert the frozen model of mixnet-s to Core ML using the tfcoreml
converter. But when I feed the .pb
file of this model to the converter, I get the following error:
OutOfRangeError: Node 'mixnet-s/mixnet_model/stem/batch_normalization/FusedBatchNormV3' (type: 'Add', num of outputs: 1) does not have output 5
I think I need to restore the model with the build function's argument training = False
and then freeze the entire graph. But I'm completely unsure about this approach. Though I'll keep looking, please let me know if you have any suggestions. I'm relatively new to TF.
Thanks in advance!
Hi!
I'm trying to freeze the graph for mixnet-s. But I'm not sure how to get the output node names. Please let me know.
Thanks!