Skip to content

Conversation

lucienwang1009
Copy link
Collaborator

  1. correct input dtype when run pretrained model.
  2. support more dtype for Maximum and Minimum and add unittests.
  3. any empty input of concat op will lead to seg fault on onnxruntime. here I workaround it by remove concat op when its input is empty.
  4. fix bugs when parsing tensor value of const op.

@lucienwang1009 lucienwang1009 force-pushed the pr_for_new_model branch 3 times, most recently from 844348f to c4817cf Compare March 22, 2019 03:09
@nbcsm
Copy link
Collaborator

nbcsm commented Mar 25, 2019

@lucienwang1009 , please add a few tests to cover these special cases in tf_to_onnx_tensor and get_tf_tensor_data, maybe add to test_internals?
either in this pr or a new one.

@nbcsm
Copy link
Collaborator

nbcsm commented Mar 25, 2019

@lucienwang1009, and also a test to cover concat input removal case.

@nbcsm nbcsm merged commit c6e55d6 into onnx:master Mar 26, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants