Can't run train.py

[Whooptie]

when i try to run train.py after the data is prepaird ik get this problem:

========= RESTART: M:\AI\nmt-chatbot-master\nmt-chatbot-master\train.py ========
Traceback (most recent call last):
File "M:\AI\nmt-chatbot-master\nmt-chatbot-master\train.py", line 8, in
from nmt import nmt
File "M:\AI\nmt-chatbot-master\nmt-chatbot-master\nmt\nmt.py", line 28, in
from . import inference
File "M:\AI\nmt-chatbot-master\nmt-chatbot-master\nmt\inference.py", line 25, in
from . import gnmt_model
File "M:\AI\nmt-chatbot-master\nmt-chatbot-master\nmt\gnmt_model.py", line 261, in
class GNMTAttentionMultiCell(tf.nn.rnn_cell.MultiRNNCell):
AttributeError: module 'tensorflow._api.v2.nn' has no attribute 'rnn_cell'

pleas help i'm a verry new newby to python

[dusty-dusty]

#143

[Nathan-Chell]

You are using Tensorflow 2.x. for this code you need to use Tensorflow 1.x preferably 1.15. You'll have to uninstall tf 2 and explicitly install tf 1.15 with: pip install tensorflow-gpu==1.15

[Whooptie]

I did install tensorflow 1.14 yesterday and it all worked, but wen i started training again it filles the shuffle and then its terminated, any idea why? Btw i am only using cpu its a ryzen 5 1600, its running in ubuntu in virtualbox

[Nathan-Chell]

What do you mean by terminated? Any errors ?

My best guess is a OOM error

[Whooptie]

no errors at all

start step 0, lr 0.001, mon dec 21 13:30:47 2020

Init train iterator, skipping 0 elements

2020-12-21 13:30:58.275320: I tensorflow/cor/kernels/shuffle_dataset_op.cc:110] Filling up shuffle buffer (this may take a while): 101395 of 128000
2020-12-21 13:31:00.976069: I tensorflow/cor/kernels/shuffle_dataset_op.cc:121] shuffle buffer filled
terminated

then the whole system freezes and training stopped

[M4A1SD]

You are using Tensorflow 2.x. for this code you need to use Tensorflow 1.x preferably 1.15. You'll have to uninstall tf 2 and explicitly install tf 1.15 with: pip install tensorflow-gpu==1.15

this pip doesnt work
i tried downloading manually [ https://pypi.org/project/tensorflow-gpu/1.15.0/#files ]

is there another way?

[Nathan-Chell]

You are using Tensorflow 2.x. for this code you need to use Tensorflow 1.x preferably 1.15. You'll have to uninstall tf 2 and explicitly install tf 1.15 with: pip install tensorflow-gpu==1.15

this pip doesnt work i tried downloading manually [ https://pypi.org/project/tensorflow-gpu/1.15.0/#files ]

is there another way?

You are using python 3.10 which no longer supports tensorflow 1.x. You need to uninstall it and download python 3.7 - or you can create a new virtual environment if you have gone that route.