Use mlc logger in scripts instead of print statements

[arjunsuresh]

We now have a global logger from MLC being used in Script automation. We need to modify the individual scripts also to use this logger instead of print statements.

[sujik18]

@arjunsuresh, do we need to similarly modify all the scripts present inside the script folder individually to replace the print statements with the usage of logger?

[arjunsuresh]

@sujik18 That's correct but a small change is needed. If we initialise logging in each file, they'll log independently which is not desirable. We do have an option to get the global logger in customize.py by doing the following

logger = i['automation'].logger

and then we can do

logger.warning('CM warning: {}'.format(e))

which uses the global logger.

[sujik18]

---

print("")
print("Data layout: {}".format(data_layout))
print("Input layers: {}".format([str(x) for x in sess.get_inputs()]))
print("Output layers: {}".format([str(x) for x in sess.get_outputs()]))
print("Input layer name: " + input_layer_name)
print("Expected input shape: {}".format(model_input_shape))
print("Output layer name: " + output_layer_name)
print("Data normalization: {}".format(normalize_data_bool))
print("Subtract mean: {}".format(subtract_mean_bool))
print('Per-channel means to subtract: {}'.format(given_channel_means))
print("Background/unlabelled classes to skip: {}".format(bg_class_offset))
print("")

---

@arjunsuresh if there are print statements which might be useful for both user and debugging purpose like the above, in such case is it better to implement both by initializing these messages in an array and then executing using a loop to implement both print and logger statements.

---

info_messages = [
"",
"Data layout: {}".format(data_layout),
"Input layers: {}".format([str(x) for x in sess.get_inputs()]),
"Output layers: {}".format([str(x) for x in sess.get_outputs()]),
"Input layer name: " + input_layer_name,
"Expected input shape: {}".format(model_input_shape),
"Output layer name: " + output_layer_name,
"Data normalization: {}".format(normalize_data_bool),
"Subtract mean: {}".format(subtract_mean_bool),
'Per-channel means to subtract: {}'.format(given_channel_means),
"Background/unlabelled classes to skip: {}".format(bg_class_offset),
""
]

for message in info_messages:
print(message)
logger.info(message)

---

Or else, is it the case that we have to strictly replace all the print statements with logger in the scripts and have to remove the print statements entirely?

[arjunsuresh]

@sujik18 These messages are in custom python files right? I think we need not change them to logger functions. We only need to change them in customize.py file which is the only python file directly run by MLC.