Description
Describe the issue
We are trying to register the model with pzmm sasctl (1.10.1) and MAS in Viya 3.5. When we try to score, it fails with the error: "EM_PREDICTION = prediction[0][0] TypeError: 'float' object is not subscribable" Error: pm.execute failed. resultCode= -1958744061
So far, we found out that if the input data shape =1, then the result must be a scalar value prediction[0][0]. Our model returns an array of floats and fails because of this.
Is it possible to have a model input as a row but the output as a table?
To Reproduce
Attaching notebook and SAS MM log file from GUI score button.
MAS_PYMAS_ERROR_FLOAT.txt
register_model.txt
Expected behavior
When I use a single row to score as a dictionary using the below code it should return a predicted value:-
from sasctl.services import microanalytic_score as mas
score_obj = dict(data_to_score.iloc[0,:][slump_col])
mas_obj = mas.execute_module_step("slump0min", "score", **score_obj)
for key, value in mas_obj.items():
print(f"{key}: {value}")
resultCode: -1958743884.0
EM_PREDICTION: None
When I use the score GUI from SAS MM using a data table stored in PUBLIC lib with more than one row, it should give me the scored regression output from catboost model
Stack Trace
ERROR: Line 301: Traceback (most recent call last):
File "/opt/sas/viya/home/SASFoundation/misc/embscoreeng/mas2py.py", line 942, in invoke
out = up[1].get(func)0(*args)
File "/tmp/tmprluhpq0f/model_exec_932ed736-60cf-4a85-8edf-4097d9741c44.py", line 55, in score
EM_PREDICTION = prediction[0][0] TypeError: 'float' object is not subscribable
Version
sasctl (1.10.1)
