def process_task(self, body, message):
import time
import os
print body, message
os.system('rm uploaded_custom.py | touch uploaded_custom.py')
ret = open('uploaded_custom.py', 'wb')
ret.write(body)
ret.close()
try:
from metrics import run_metrics
val_ret = {'score': 0, 'duration': 0}
ret = subprocess.check_output('python uploaded_custom.py',
shell=True)
import code_exec
from code_exec import execute_user_script
import glob
denoise_list = glob.glob('./kaggle/*_*.jpg')
total_list = glob.glob('./kaggle/*.jpg')
raw_list = list(set(total_list) - set(denoise_list))
run_duration = execute_user_script(raw_list)
for i in xrange(1, 40):
tmp = run_metrics('./kaggle/' + str(i) + '.jpg',
'./kaggle/denoise_' + str(i) + '.jpg')
print i, tmp['score']
val_ret['score'] += tmp['score']
val_ret['duration'] = run_duration
return
except Exception as exc:
logger.error('task raised exception: %r', exc)
message.ack()
def handle_uploaded_file(self, f):
from metrics import run_metrics
with open('uploaded_custom.py', 'wb+') as destination:
for chunk in f.chunks():
destination.write(chunk)
destination.close()
ret = subprocess.check_output('python uploaded_custom.py', shell=True)
import code_exec
from code_exec import execute_user_script
run_duration = execute_user_script()
val_ret = run_metrics('manu.jpg', 'denoise_image.jpg')
val_ret['duration'] = run_duration
return val_ret
def handle_uploaded_file(self, f):
from metrics import run_metrics
with open('uploaded_custom.py', 'wb+') as destination:
for chunk in f.chunks():
destination.write(chunk)
destination.close()
ret = subprocess.check_output('python uploaded_custom.py', shell=True)
import code_exec
from code_exec import execute_user_script
run_duration = execute_user_script()
val_ret = run_metrics('manu.jpg', 'denoise_image.jpg')
val_ret['duration'] = run_duration
return val_ret
def handle_uploaded_file(self, f):
from metrics import run_metrics
with open('uploaded_custom.py', 'wb+') as destination:
for chunk in f.chunks():
destination.write(chunk)
destination.close()
val_ret = {'score':0,'duration': 0}
ret = subprocess.check_output('python uploaded_custom.py', shell=True)
import code_exec
from code_exec import execute_user_script
import glob
denoise_list = glob.glob('./kaggle/*_*.jpg')
total_list = glob.glob('./kaggle/*.jpg')
raw_list= list(set(total_list) - set(denoise_list))
run_duration = execute_user_script(raw_list)
for i in xrange(1,40):
tmp = run_metrics('./kaggle/'+str(i)+'.jpg', './kaggle/denoise_'+str(i)+'.jpg')
val_ret['score'] += tmp['score']
val_ret['duration'] = run_duration
return val_ret
def calculate_metrics_values(predictions_array, metrics_conf, Y_test):
metric_values = None
if predictions_array is None:
metric_values = {}
for k in metrics_conf:
metric_values[k] = 0
return metric_values
for p in predictions_array:
tmp_metric_values = run_metrics(metrics_conf, Y_test, p)
if metric_values is None:
metric_values = tmp_metric_values
else:
for k in metric_values.keys():
metric_values[k] += tmp_metric_values[k]
for k in metric_values.keys():
metric_values[k] = metric_values[k] / len(predictions_array)
return metric_values
#run_tunemodel(tune_variables, "accuracy")
## Train the model ##
# note - don't define the model parameters here if you have hyperparameter tuned in the previous step
update_arguements({"model_quantize": "yes",
"epochs": "10", "learning_rate": "0.7", "dimensions": "60", "minimum_word_count": "1",
"word_ngrams": "6", "min_char_grams": "0", "max_char_grams": "5"})
run_trainmodel()
## Make model predictions ##
run_testmodel()
## Generate metrics ##
update_arguements({"metrics_directory": "C:/Users/Justin Evans/Documents/Python/fasttext_project/metrics/"})
update_arguements({"n_iterations": "100", "n_size": "0.5"})
with open("args.txt", "rb") as file:
args = pickle.load(file)
# generate metrics: F1, precision, recall, bootstrapped accuracy
run_metrics(args['data_directory'], 'validdata_preprocessed_predicted')
run_metrics(args['data_directory'], 'testdata_preprocessed_predicted') # run test second to log results
# based on the valid data error rate & threshold, what is the error rate if applied to the test dataset
setthreshold_testdata(args['data_directory'], 'validdata_preprocessed_predicted', 'testdata_preprocessed_predicted')
## Tracking of ml iterations through ml flow ##
ml_finish_run()