The clear_session() method of keras.backend does not clean up the previously execution

[ram]

When I run the model, I find that it was conflicting with the previously executed model. In addition,  global variables in my program contain values from the previous run. I used keras.backend.clear_session() and del model but the problem persists. How can I return to the initial state at each execution?
I used google colab notebook as an execution environment with keras version 2.3.1 and Tensorflow backend 2.2.0

[Lance Norskog]

Use the Colab 'restart runtime' option. 

Lance

On Fri, Apr 24, 2020 at 5:07 PM ram <ram...@gmail.com> wrote:

When I run the model, I find that it was conflicting with the previously executed model. In addition,  global variables in my program contain values from the previous run. I used keras.backend.clear_session() and del model but the problem persists. How can I return to the initial state at each execution?
I used google colab notebook as an execution environment with keras version 2.3.1 and Tensorflow backend 2.2.0

--
You received this message because you are subscribed to the Google Groups "Keras-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keras-users...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/d30614f1-7e13-4097-a0f0-a7d09af9fd1f%40googlegroups.com.

--

Lance Norskog
lance....@gmail.com
Redwood City, CA

[ram]

Thanks Lance,

 If I use a loop in my program for training many models. How I can do for each iteration to reset the model to the "initial" state for training another one?

 Can i use reset_seeds() at the start of each loop in addition del model and keras.backend.clear_session() at the end ? This in order to clear the memory. 

[Lance Norskog]

Ah, ok. Yes, global variables are retained as long as the notebook runtime is up. There is a 'gc' package in Python to call Python's garbage collection. I have used this in the past (not with keras).

Colab has a UI item in the top right corner that shows you how much ram is used in CPU and GPU, and disk space. 

Lance

--

You received this message because you are subscribed to the Google Groups "Keras-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keras-users...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/2b686dd3-fba7-4af1-a916-8d8092235cdc%40googlegroups.com.