How to distinguish between stock TensorFlow and Intel TensorFlow for 2.5.0 version quickly and conveniently.

[zhixingheyi-tian]

How to distinguish between stock TensorFlow and Intel TensorFlow for 2.5.0 version quickly and conveniently.

@dmsuehir @ashahba Would you please guide this?

[dmsuehir]

@zhixingheyi-tian The documentation here has a section called "sanity check" with a sample script on how to do this. Using info from that script with pip install tensorflow==2.5.0 and pip install intel-tensorflow==2.5.0 I found that this worked:

from tensorflow.python.util import _pywrap_util_port
_pywrap_util_port.IsMklEnabled()

It returns True for intel-tensorflow and False for stock tensorflow.

[zhixingheyi-tian]

@dmsuehir
Thanks very much.

For Stock TensorFlow 2.5.0, if enable os.environ["TF_ENABLE_ONEDNN_OPTS"] = '1', does it mean completely equivalent to Intel TensorFlow 2.5.0 , and does it include all optimizations from Intel TensorFlow?

Thanks

[dmsuehir]

@zhixingheyi-tian There's a table in the documentation here under the section called "Differences between Intel Optimization for Tensorflow and official TensorFlow for running on Intel CPUs after v2.5" that compares Intel TensorFlow to stock TensorFlow. It also shows which features are enabled with TF_ENABLE_ONEDNN_OPTS. You'll be able to see from there that it's not completely equivalent.

[PLEX-GR00T]

Does this thing work in Google Colab? I was trying pip install intel-tensorflow in the google colab, and it also shows True for
from tensorflow.python.util import _pywrap_util_port'
_pywrap_util_port.IsMklEnabled()

But, Ido not have GPU access of the Colab afterwards why? what should I do. It is more slower then usual Colab GPU speed.

[sramakintel]

@PLEX-GR00T and @zhixingheyi-tian is your issue resolved?