time_budget = max_time
overall_time_budget = overall_time_budget + time_budget
vprint( verbose, "[+] Cumulated time budget (all tasks so far) %5.2f sec" % (overall_time_budget))
# We do not add the time left over form previous dataset: time_budget += time_left_over
vprint( verbose, "[+] Time budget for this task %5.2f sec" % time_budget)
time_spent = time.time() - start
vprint( verbose, "[+] Remaining time after reading data %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint( verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
# ========= Creating a model, knowing its assigned task from D.info['task'].
# The model can also select its hyper-parameters based on other elements of info.
vprint( verbose, "======== Creating model ==========")
M = MyAutoML(D.info, verbose=False, debug_mode=debug_mode) # I turned off verbose to avoid tons of junk...
print M
# ========= Iterating over learning cycles and keeping track of time
time_spent = time.time() - start
vprint( verbose, "[+] Remaining time after building model %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint( verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
time_budget = time_budget - time_spent # Remove time spent so far
start = time.time() # Reset the counter
time_spent = 0 # Initialize time spent learning
cycle = 0
while time_spent <= time_budget/2 and cycle <= max_cycle and M.model.n_estimators<max_estimators:
time_budget = float(time_budget)
overall_time_budget = overall_time_budget + time_budget
vprint(verbose, "[+] Cumulated time budget (all tasks so far) %5.2f sec" % (overall_time_budget))
# We do not add the time left over form previous dataset: time_budget += time_left_over
vprint(verbose, "[+] Time budget for this task %5.2f sec" % time_budget)
time_spent = time.time() - start
vprint(verbose, "[+] Remaining time after reading data %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint(verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
# ========= Creating a model, knowing its assigned task from D.info['task'].
# The model can also select its hyper-parameters based on other elements of info.
vprint(verbose, "======== Creating model ==========")
M = MyAutoML(D.info, verbose=False, debug_mode=debug_mode, num_train=num_train, num_valid=num_valid, num_test=num_test, dim_data=dim_data)
print(M)
# ========= Iterating over learning cycles and keeping track of time
time_spent = time.time() - start
vprint(verbose, "[+] Remaining time after building model %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint(verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
time_budget = time_budget - time_spent # Remove time spent so far
start = time.time() # Reset the counter
time_spent = 0 # Initialize time spent learning
cycle = 0
]
D.info['hyps'] = _convert_to_hyp(space_hyps, base_hyps=[is_bo])
arr_range = np.array([[elem_hyps[3], elem_hyps[4]]
for elem_hyps in space_hyps])
model_bo = bayesobo.BO(arr_range, str_acq='ucb')
cur_hyps = [elem_hyps[2] for elem_hyps in space_hyps]
list_hyps_all = []
list_measures_all = []
while time_spent <= time_budget - (
2 * time_one_loop) - TIME_RESIDUAL and is_bo and True:
vprint(
verbose, "=========== " + basename.capitalize() +
" Training cycle " + str(cycle) + " ================")
M = MyAutoML(D.info, verbose=False, debug_mode=debug_mode)
M.fit(X_train_train, Y_train_train)
vprint(
verbose, "[+] Fitting success, time spent so far %5.2f sec" %
(time.time() - start))
# Make predictions
# -----------------
pred_train_valid = M.predict(X_train_valid)
if 'classification' in D.info['task']:
performance = sklearn.metrics.roc_auc_score(
Y_train_valid, pred_train_valid)
performance = 2 * performance - 1
vprint(verbose,
"[+] AUC for X_train_valid, %5.4f" % (performance))
if time_spent >= time_budget:
vprint( verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
# ========= Creating a model, knowing its assigned task from D.info['task'].
# The model can also select its hyper-parameters based on other elements of info.
vprint( verbose, "======== Creating model ==========")
if D.info['task'] !='regression':
import engine_automl as mpd
mpd.predict(D, output_dir, start, time_budget, basename, running_on_codalab )
else:
M = MyAutoML(D.info, verbose, debug_mode)
# ========= Iterating over learning cycles and keeping track of time
# Preferably use a method that iteratively improves the model and
# regularly saves predictions results gradually getting better
# until the time budget is exceeded.
# The example model we provide we use just votes on an increasingly
# large number of "base estimators".
time_spent = time.time() - start
vprint( verbose, "[+] Remaining time after building model %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint( verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
time_budget = time_budget - time_spent # Remove time spent so far
overall_time_budget = overall_time_budget + time_budget
vprint( verbose, "[+] Cumulated time budget (all tasks so far) %5.2f sec" % (overall_time_budget))
# We do not add the time left over from previous dataset: time_budget += time_left_over
vprint( verbose, "[+] Time budget for this task %5.2f sec" % time_budget)
time_spent = time.time() - start
vprint( verbose, "[+] Remaining time after reading data %5.2f sec" % (time_budget-time_spent))
if time_spent >= time_budget:
vprint( verbose, "[-] Sorry, time budget exceeded, skipping this task")
execution_success = False
continue
# ========= Creating a model, knowing its assigned task from D.info['task'].
# The model can also select its hyper-parameters based on other elements of info.
# Only instantiates class object
vprint( verbose, "======== Creating model ==========")
M = MyAutoML(D.info, verbose=False, debug_mode=debug_mode) # I turned off verbose to avoid tons of junk...
print M
# Does cross validation for all models and picks the best classifier
# Probably need to pass in the time remaining since timekeeping needs to be done in here
# -------------------
M.run_cycles(D.data['X_train'], D.data['Y_train'])
vprint( verbose, "[+] Fitting success, time spent so far %5.2f sec" % (time.time() - start))
vprint( verbose, "[+] Size of trained model %5.2f bytes" % data_io.total_size(M))
# Make predictions
# -----------------
Y_test = M.predict(D.data['X_test'])
vprint( verbose, "[+] Prediction success, time spent so far %5.2f sec" % (time.time() - start))
# Write results