def main(job_id, params):
job_id = str(job_id)
print(params)
# Encode params to pass it through the run_command
my_params = {}
for key in params.keys():
my_params[key.replace('"', '\'')] = int(params[key])
my_params = json.dumps(my_params)
dico_to_save = {}
dico_to_save['platform'] = 'Movidius'
dico_to_save['job'] = job_id
dico_to_save['params'] = my_params
#Create a unique identifier for this model
name = f.create_name(my_params)
if cfg.debug:
max_number_of_steps = 100
else:
max_number_of_steps = 28125 #With a batch_size at 32, we train for 20 epochs
#Launch training of the model on the cloud
f.train_on_cloud(name, job_id, public_DNS, instance_id, nb_gpus_to_use,
gpus_to_use, max_number_of_steps, dico_to_save, ID_NUMBER)
#Test the trained model on the platform
dico_hardware = f.test_movidius(name, max_number_of_steps, job_id)
return dico_hardware
def main(job_id, params):
job_id = str(job_id)
print(params)
# Encode params to pass it through the run_command
my_params = {}
for key in params.keys():
my_params[key.replace('"', '\'')] = int(params[key])
my_params = json.dumps(my_params)
dico_to_save = {}
dico_to_save['platform'] = 'Movidius'
dico_to_save['job'] = job_id
dico_to_save['params'] = my_params
name = f.create_name(my_params)
if cfg.debug:
max_number_of_steps = 100
else:
max_number_of_steps = 28125
train_function(name, job_id, nb_gpus_to_use, gpus_to_use,
max_number_of_steps, dico_to_save, ID_NUMBER)
dico_hardware = f.test_movidius(name, max_number_of_steps, job_id)
return dico_hardware