-
Notifications
You must be signed in to change notification settings - Fork 26
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Enquiry about parsing optional input #22
Comments
Hi @erizmr thank you for reporting this issue. This is really a good example of the operator semantic divergence between TFLite and ONNX. I checked the
In tflite2onnx, we may have a helper to create empty tensor and make it as inputs of an operator. The challenge here is, graph level handling (e.g. data layout handling) may be impacted as we seem to assume tensors are not empty. Do you have a simplified TFLite model that contains the |
Thansk for your reply. @jackwish
There are two inputs: Maybe we can start with this simple model. |
@erizmr thank you for the update. May I know which version of the TensorFlow you are using to generate this TFLite model? It's interesting that the Background: any attributes in TFLite model shall have a reader method in its parser, e.g. |
That's the definition of TensorFlow operator which is not always same as its TFLite peer. |
Hi @jackwish, I tried the empty tensors(from Maybe I can send a pull request to share the code if you wish. |
Cool! A PR would be great! Thanks for the update @erizmr |
Hi jackwish,
Thanks for sharing your awesome frame!
I am trying to enable a new operator
Resize
fromresize_bilinear
, the ONNX description ofResize
is shown below:Only one of 'scales' and 'size' should be specified in this case, I am not sure how to 'skip' one of the arguments (also the
roi
) when implementing the parse part, because it seems the inputs are store in a list without a key.I am wondering if you could share any suggestions about it?
Thanks,
Mingrui
The text was updated successfully, but these errors were encountered: