Rémi Cresson

``` cresson@cin-mo-gpu:~/decloud$ docker run -ti mdl4eo/otbtf2.4:cpu bash -c "python -c 'import tensorflow; print(tensorflow.__version__)'" 2.4.1 ``` I don't understand why the image I push on dockerhub is not the right one!

Weird. It looks like I pushed the good images. But I am unable to use them on my (intel) laptop, even though they are running fine on the gpu servers....

Server 1 (working): ``` Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian CPU(s): 8 On-line CPU(s) list: 0-7 Thread(s) per core: 1 Core(s) per socket: 4 Socket(s): 2...

Thanks for this analysis... so you were right about bazel optimization flags.

So for building docker images that work on most computers, better sticking to basic optimization flags...

Thanks. This was a bit off topic, but clearly deserved some explanation. And one less issue for windows users...

Hi @viniciuspg , I can't see your screen capture :open_mouth: I never tried **docker desktop** myself, but it looks like you can use the Windows **PowerShell** to run docker images...

Hello @farissetiawan6 The loss value is implemented in the network graph. For instance if you take a look in [this example](https://github.com/remicres/otbtf/blob/develop/otbtf/examples/tensorflow_v1x/create_savedmodel_simple_cnn.py#L69) you can see the loss. Evaluation metrics are another...

The graph is a concept in TensorFlow. You can check the TensorFlow documentation for more info

There was an error in the documentation: wrong patches sizes during training step (I just corrected it in https://github.com/remicres/otbtf/commit/3daef29fa0a741497d47abc13b83829434ae511f). You must change the patches sizes for input and output, because...