Hi all, I'm working on a predictive modeling pipeline in SAS Viya using Open Source Code nodes (Python). The pipeline works well during training, but when I add a Scoring Node, I get this : ERROR: Scoring cannot be completed because new variables have been created by one or more Open Source node(s). I understand this may be due to the fact that new variables (like predictions) are being created dynamically in the Python code, but I haven't found a clear solution to properly register or declare them for scoring. I've had several long chats with ChatGPT — no luck so far, I'm now in desperate need of help I'm attaching the code below for context. Any help, advice, or working example would be sincerely appreciated! Thanks in advance 🙏 Training Code : import pandas as pd
import numpy as np
import lightgbm as lgb
target_col = 'Stress_level'
target_months = [202308, 202309, 202409, 202410]
record = dm_inputdf.copy()
dm_interval_input = ["DATA_YM", "blood_pressure", "sleep_duration", "work_hours", "age"]
rec_intv = record[dm_interval_input].astype(np.float32)
rec_all = pd.concat([rec_intv.reset_index(drop=True)], axis=1)
record["target_group"] = np.where(record["DATA_YM"].isin(target_months), "tr1", "tr2")
tr1 = rec_all[record["target_group"] == "tr1"]
tr2 = rec_all[record["target_group"] == "tr2"]
y_tr1 = record.loc[record["target_group"] == "tr1", target_col].astype(np.float32)
y_tr2 = record.loc[record["target_group"] == "tr2", target_col].astype(np.float32)
model_tr1 = lgb.LGBMRegressor(
n_estimators=3000,
learning_rate=0.05,
max_depth=12,
num_leaves=13,
force_row_wise=True
)
model_tr1.fit(tr1, y_tr1)
model_tr2 = lgb.LGBMRegressor(
n_estimators=3000,
learning_rate=0.05,
max_depth=12,
num_leaves=13,
force_row_wise=True
)
model_tr2.fit(tr2, y_tr2)
pred_tr1 = model_tr1.predict(tr1)
pred_tr2 = model_tr2.predict(tr2)
record.loc[record["target_group"] == "tr1", "P_Stress_level"] = np.clip(pred_tr1, 0, None)
record.loc[record["target_group"] == "tr2", "P_Stress_level"] = np.clip(pred_tr2, 0, None)
dm_scoreddf = record.copy()
dm_scoreddf["P_Stress_level"] = dm_scoreddf["P_Stress_level"].astype(np.float64)
dm_scoreddf=dm_scoreddf[[
"Stress_level", "blood_pressure", "sleep_duration", "work_hours", "age", "P_Stress_level", "DATA_YM"]]
dm_scoreddf["P_Stress_level"].attrs.update({
"role": "PREDICTION",
"level": "INTERVAL",
"description": "LightGBM Predition"
}) Scoring Code : import pandas as pd
import numpy as np
def score_method(blood_pressure, sleep_duration, work_hours, age, DATA_YM):
"Output: P_Stress_level"
record = pd.DataFrame([[blood_pressure, sleep_duration, work_hours, age, DATA_YM
]], columns=['blood_pressure', 'sleep_duration', 'work_hours', 'age', 'DATA_YM'
])
dm_interval_input = [col for col in record.columns if col not in dm_class_input]
rec_intv = record[dm_interval_input]
rec_intv_imp = imputer.transform(rec_intv)
rec = np.concatenate((rec_intv_imp), axis=1)
rec_pred = model_tr1.predict(rec) if int(DATA_YM) in target_months else model_tr2.predict(rec)
return float(np.clip(rec_pred[0], 0, None))
... View more
