tracking, printing, and graphing log loss and accuracy

[evil242]

I want to thank you for putting this out. It is helping to both learn tensorflow and inception. Is there a modification you would recommend to save log loss and accuracy to the summary dir?

I'm currently working to upgrade this to python3.5 w/ tensorflow 1.4 (some functions have been deprecated or eliminated). And then apply some recommendations from Martin Gorner's learning tensorflow without a PhD module.

[YuntaoTan]

python3 decode error

@evil242 , I am very happy, recently, i am doing the same work as you, i use python3.4. Firstly, I use 2to3 get the hint of translation python2 script to python3. when i run " python3 -m src.data_preparation.stanford_ds_to_tfrecord ", but i get the error "UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte"; I know this is decode problem, but , i dont know how to dubug it. I am a newer for python. can you help me ?

[evil242]

For all of the open commands, in addition to the 'r' read flag, you need to add the 'b' - byte flag open(path, 'rb').read()

[YuntaoTan]

@evil242 ,thank you very much, that's right!