Help me!

[buicongthek58]

Hi guy!
I tried run your code with python 2.7. and docker, it's run very good. Can you tell me why this source must run with this environment?
Now, I want to run this code with python 3 and without docker. How can I do that?
Thanks a lot!

[amurzina]

Hi!

1. you need to install all the requirements for python3.*
2. and follow the exceptions you've got, you need to fix compatibility errors: python2 style -> python3 style.

[buicongthek58]

I tried run with python 3. But the results this so different than python 2 and docker
Step 0 (epoch 1), average_train_loss = 4.64211, step_loss = 4.64211, time_per_step = 1.017 Step 1 (epoch 2), average_train_loss = 1.03712, step_loss = 0.33868, time_per_step = 1.953 Step 2 (epoch 4), average_train_loss = 0.62684, step_loss = 0.14512, time_per_step = 1.928 Step 3 (epoch 5), average_train_loss = 0.45254, step_loss = 0.07919, time_per_step = 2.058 Step 4 (epoch 7), average_train_loss = 0.35200, step_loss = 0.03626, time_per_step = 1.976

And TP rate, Precision, Recall is 0

[buicongthek58]

I found the reason why:

vocab.py:
try: text = text.decode('utf-8') characters = list(text) except Exception as e: characters = ['<UNK>']

In python3: not need text = text.decode('utf-8')
Because it has exception so I can't find that :))
Thanks a lot!

[Tikam02]

@buicongthek58 so what to do.
should i remove that line.

[RutujaWanjari]

@buicongthek58 , @Tikam02 @amurzina , any luck on his front?

I have also converted whole code to 3.6. Notebook is running fine, but Recall and precision are 0.0

[amurzina]

We don't support this version of POC any more.
I can just give a suggestion to fit the model again and fix hyperparameters.