TensorFlow r1.4 | Hacker News Reader

TensorFlow r1.4

developers.googleblog.com

103 pointsplexicle13 comments

Still, the old issue persists, just issue a "import tensorflow as tf" with python3 you got:

    ~tensorflow/lib/python3.6/importlib/_bootstrap.py:219: 
    RuntimeWarning: compiletime version 3.5 of module 
    'tensorflow.python.framework.fast_tensor_util' does not 
    match runtime version 3.6
    
    return f(*args, **kwds)

This does not prevent me from running short code, however who knows what this potentially means, better fixing that before the release. It obviously said the runtime was built with 3.5 instead of 3.6.

https://github.com/tensorflow/tensorflow/issues/14182

It's just a small Cython-compiled utility throwing the warning, it works fine. Everything else is pure Python and so doesn't care about 3.5 vs 3.6. We will fix the warning, though (disclaimer: I work on TF and added fast_tensor_utils--sorry!).

How do you import Keras from TensorFlow instead of from the Keras standalone itself? Just substitute tf.keras where appropriate?

yep

From tensorflow import keras

Trivially simple example here: https://www.kaggle.com/dansbecker/transfer-learning-scratch-...

    # Use pip for Python 2.7
    # Use pip3 instead of pip for Python 3.x

Why would Google help to prolong the life of Python 2.x? They usually seem so interested in pushing technologies forward.

I think Python 2.x is like XP/IE. It's supported so that more people can quickly benefit from the framework, which ultimately pushes the technology forward.

Nowadays there are very few statistical packages which only support Python 2 and not Python 3.

that ver few packages nailed most people on python 2 boat.

Internally Google still uses Python 2.7 only.

From my understanding this is because the new projects use golang or java.

https://www.reddit.com/r/Python/comments/3cak9z/why_does_goo...

Those information is not accurate. First, Google has been trying to migrate to Python 3 for years, but the implementation of the plan has been delayed all the time. Maybe this year certain internal codes will finally be migrated fully. Second, many new projects are still in Python. Even if some Googlers hate Python, they have no other choice if they want to do AI.

I was sad to see `tf.contrib.data.get_single_element()` didn't make it in :(

Keep building TF Serving from source until the next release :(

Reply on news.ycombinator.com