You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/06/24 10:50:06 UTC
[GitHub] [tvm] rafzi opened a new pull request, #11877: [Frontend][TFLite] respect out type of Shape op
rafzi opened a new pull request, #11877:
URL: https://github.com/apache/tvm/pull/11877
This issue popped up when reading a model that uses the output of `Shape` with `out_type=int32` and concatenating with other int32s. The Shape output was of type int64, so the concatenation reported an error in type inference.
I'm not sure why this test does not catch the issue. Should it not also verify that the type was really set to int32?
https://github.com/apache/tvm/blob/main/tests/python/frontend/tflite/test_forward.py#L1771
@dhruvaray @siju-samuel @anijain2305
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] rafzi commented on pull request #11877: [Frontend][TFLite] respect out type of Shape op
Posted by GitBox <gi...@apache.org>.
rafzi commented on PR #11877:
URL: https://github.com/apache/tvm/pull/11877#issuecomment-1178261223
@AndrewZhaoLuo @hlu1 @jwfromm
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] rafzi commented on pull request #11877: [Frontend][TFLite] respect out type of Shape op
Posted by GitBox <gi...@apache.org>.
rafzi commented on PR #11877:
URL: https://github.com/apache/tvm/pull/11877#issuecomment-1181604943
@AndrewZhaoLuo Thank you for looking at this. I have updated the test accordingly. I had to add a `tf.add` to make it fail without the fix.
The int64 default comes from a different `shape_of` function in the frontend code, not from `relay.shape_of`:
https://github.com/apache/tvm/blob/fc419df32f052e21f614c8940699c10a2d696689/python/tvm/relay/frontend/common.py#L612
`tf.shape` only supports int32 and int64: https://www.tensorflow.org/api_docs/python/tf/shape#args
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] AndrewZhaoLuo merged pull request #11877: [Frontend][TFLite] respect out type of Shape op
Posted by GitBox <gi...@apache.org>.
AndrewZhaoLuo merged PR #11877:
URL: https://github.com/apache/tvm/pull/11877
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org