Closed MarkTension closed 4 years ago
Aah it seems that there was a problem with tensorflow 1.15 , and an older version of tensorflow_to_barracuda.py according to https://github.com/Unity-Technologies/ml-agents/issues/2775 . Updated the script to the latest version, and it works:)
I am not able to convert my .pb file to bytes using any of the above suggestions. My tf version is 1.15,,, request you to please guide which git branch i should take,,, have tried with master, 0.15 already,,
I am getting the same "Assertion Error " as detailed above
Maybe take that to the mlagents unity forum and specify what command, and kind of model. Custom model, or ml-agent generated etc. G'luck
@ankur-00007 We haven't used .bytes
files in a long time; I believe that was for using tensorflow sharp. You should refer to the documentation for the version of ML-Agents that you're using, or upgrade to the latest version and follow its documentation.
This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.
hi all,
when using tensorflow_to_barracuda.py , with either a mobilenet_v2.pb or inception.pb model file I get either
assert x.data.size == np.prod(shape) AssertionError
with inception,
or
assert all_elements_equal(input_ranks) AssertionError
with mobilenet_v2
I guess this is because of unsupported operations? If so, are there any other recommended models for image classification that don't have unsupported operations? Or any recommended ways to remove them?
Or maybe less problems with TensorflowSharp instead of barracuda?
Thanks!