kwargs.update(json.loads(params))
kwargs[
'model_id'] = 'DSS.H2O_connector.model.' + output_folder.full_name + '.' + algorithm
algorithms = {
'autoencoder': h2o.h2o.autoencoder,
'deeplearning': h2o.h2o.deeplearning,
'gbm': h2o.h2o.gbm,
'glm': h2o.h2o.glm,
'glrm': h2o.h2o.glrm,
'kmeans': h2o.h2o.kmeans,
'naive_bayes': h2o.h2o.naive_bayes,
'prcomp': h2o.h2o.prcomp,
'random_forest': h2o.h2o.random_forest,
'svd': h2o.h2o.svd,
}
# print 'Arguments passed to H2O: ', kwargs # This makes the job fail with exception None ??
model = algorithms[algorithm](**kwargs)
## save model summary in output_folder and model to disk
with open(os.path.join(output_folder_path, 'model_summary.txt'), 'w') as file:
orig_stdout = sys.stdout
sys.stdout = file
model.show() # this method uses print to write to stdout
sys.stdout = orig_stdout
h2o.h2o.save_model(model,
saved_model_folder(model_config, output_folder),
force=True) # "force" means overwrite
## load model
def find_model_id(folder):
with open(os.path.join(folder.get_path(), 'model_summary.txt')) as file:
for line in file:
match = re.match('Model Key: (DSS\.H2O_connector\.model\.'+folder.full_name+'\..*)$', line)
if match:
return match.group(1)
raise Exception('Could not find model id in model_summary.txt')
model_id = find_model_id(input_folder)
try:
model = h2o.h2o.get_model(model_id)
except EnvironmentError as e:
print 'Model key unknown to H2O:\n“',e,'”\n','Will thus load from saved model.'
path = os.path.join(saved_model_folder(model_config, input_folder), model_id)
model = h2o.h2o.load_model(path=path)
## compute predictions and add input columns to output
predict_frame = model.predict(input_frame)
print "predict_frame id:", predict_frame.frame_id
print "predict_frame types:", predict_frame.types
sys.stdout.flush()
for col in get_recipe_config()["columns_to_copy_to_output"]:
predict_frame[col] = input_frame[col]
## save output
def DSS_type(col_name):
return {
'enum':'string',
for line in file:
match = re.match(
'Model Key: (DSS\.H2O_connector\.model\.' + folder.full_name +
'\..*)$', line)
if match:
return match.group(1)
raise Exception('Could not find model id in model_summary.txt')
model_id = find_model_id(input_folder)
try:
model = h2o.h2o.get_model(model_id)
except EnvironmentError as e:
print 'Model key unknown to H2O:\n“', e, '”\n', 'Will thus load from saved model.'
path = os.path.join(saved_model_folder(model_config, input_folder),
model_id)
model = h2o.h2o.load_model(path=path)
## compute predictions and add input columns to output
predict_frame = model.predict(input_frame)
print "predict_frame id:", predict_frame.frame_id
print "predict_frame types:", predict_frame.types
sys.stdout.flush()
for col in get_recipe_config()["columns_to_copy_to_output"]:
predict_frame[col] = input_frame[col]
## save output
def DSS_type(col_name):
kwargs['model_id'] = 'DSS.H2O_connector.model.' + output_folder.full_name + '.' + algorithm
algorithms = {
'autoencoder': h2o.h2o.autoencoder,
'deeplearning': h2o.h2o.deeplearning,
'gbm': h2o.h2o.gbm,
'glm': h2o.h2o.glm,
'glrm': h2o.h2o.glrm,
'kmeans': h2o.h2o.kmeans,
'naive_bayes': h2o.h2o.naive_bayes,
'prcomp': h2o.h2o.prcomp,
'random_forest': h2o.h2o.random_forest,
'svd': h2o.h2o.svd,
}
# print 'Arguments passed to H2O: ', kwargs # This makes the job fail with exception None ??
model = algorithms[algorithm](**kwargs)
## save model summary in output_folder and model to disk
with open(os.path.join(output_folder_path, 'model_summary.txt'),'w') as file:
orig_stdout = sys.stdout
sys.stdout = file
model.show() # this method uses print to write to stdout
sys.stdout = orig_stdout
h2o.h2o.save_model(
model,
saved_model_folder(model_config, output_folder),
force=True) # "force" means overwrite
