# Import tensorflow kills jupyter notebook kernel

## post by SAICsergio on Feb 6, 2019

I am new to lambda stack. I opened a jupyter notebook from the terminal with “jupyter notebook” to test a few things. Python version seems good (“import sys”, “sys.version” > ‘3.6.7 (default, Oct 22 2018, 11:32:17) \n[GCC 8.2.0]’), and PyTorch imports properly (“import torch”, “torch. **version**” > ‘0.4.1’). But, just running “import tensorflow as tf” causes the kernel to die:

Kernel Restarting  
The kernel appears to have died. It will restart automatically.

Please help me address this. Thanks.

## post by sabalaba on Feb 8, 2019

Hi Sergio, can you explain in more detail what you mean by “Kernel Restarting” and it appearing to die. Looks like a conda / jupyter notebook issue.

This GitHub issue suggests updating conda via:

```sh
conda update anaconda
```

or alternatively using these settings for TF:

```py
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
```

## post by tensorTensor on Aug 13, 2021

Sounds just like my issue, but could be a different cause…

Same issue, now running local…

Windows10, Jupyter, Anaconda, TF installed, env active

import tensorflow as tf

popup message & subsequent red flag up top: “Dead Kernel”

Isolated via comment-out all other lines  
Import tf is singular cause: every time

in my case, found help in terminal output of import statement…  
“You can, at your own risk, disable this warning by setting the environment
variable ‘HDF5_DISABLE_VERSION_CHECK’ to a value of ‘1’.”  
This was due to “h5py is running against HDF5 1.10.5 when it was built against 1.10.6”  
Even tho I had 1.10.6 installed  
Also, found sage wisdom in all comments all the way down to the bottom here…