I am training using the Tensorflow Object Detection API in Google Colab. I want to suppress printing the loss at each step as the web page crashes after 30 minutes due to a large amount of text being printed as the output of the cell. I have to manually clear the output of the cell every 30 minutes or so to avoid this issue. Is there any way to modify the train.py code so that Tensorflow stops printing the loss at every step. I have tried changing the code in line 57 of the research/object_detection/legacy/train.py from tf.logging.set_verbosity(tf.logging.INFO) to tf.logging.set_verbosity(tf.logging.WARN) but it did not seem to work. Any suggestions/workarounds?
add ; if thats the print statement at the end of the print statement
Add %%capture in the first line of the cell for no print for the cell
and for speficic function
from IPython.utils import io
with io.capture_output() as captured:
function()