Online Evaluation: handle empty data when queried (#3593)

* handle empty data when queried

* fix incorrect span_id, trace_id

assets/evaluation_on_cloud/environments/evaluations-built-in/context/online_eval/evaluate.py

@@ -18,7 +18,7 @@ from azure.ai.evaluation import evaluate
 from azure.ai.ml.identity import AzureMLOnBehalfOfCredential
 from collections import defaultdict
 
-from utils import get_mlclient, extract_model_info
+from utils import get_mlclient, extract_model_info, is_input_data_empty
 
 logger = logging.getLogger(__name__)
 logging.basicConfig(level=logging.INFO)
@@ -228,6 +228,8 @@ rai_evaluators = [
 
 def run(args):
     """Entry point of model prediction script."""
+    if is_input_data_empty(args["preprocessed_data"]):
+        return
     evaluators = json.loads(args["evaluators"])
     # evaluators = download_evaluators_and_update_local_path(evaluators)
     evaluators = copy_evaluator_files(args)

assets/evaluation_on_cloud/environments/evaluations-built-in/context/online_eval/postprocess.py

@@ -14,6 +14,7 @@ from opentelemetry.trace.span import TraceFlags
 from opentelemetry.sdk._logs import LoggerProvider
 from opentelemetry.sdk._logs.export import BatchLogRecordProcessor
 from azure.monitor.opentelemetry.exporter import AzureMonitorLogExporter
+from utils import is_input_data_empty
 
 import logging
 
@@ -78,10 +79,10 @@ def log_evaluation_event(row) -> None:
         logger.warning("Missing required fields in the row: evaluation")
     if "trace_id" not in row:
         logger.debug(f"Missing trace_id from user query result, taking default of column {DEFAULT_TRACE_ID_COLUMN}")
-        trace_id = int(row.get("trace_id", row.get(DEFAULT_TRACE_ID_COLUMN, "0")), 16)
     if "span_id" not in row:
         logger.debug(f"Missing span_id from user query result, taking default of column {DEFAULT_SPAN_ID_COLUMN}")
-        span_id = int(row.get("span_id", row.get(DEFAULT_SPAN_ID_COLUMN, "0")), 16)
+    trace_id = int(row.get("trace_id", row.get(DEFAULT_TRACE_ID_COLUMN, "0")), 16)
+    span_id = int(row.get("span_id", row.get(DEFAULT_SPAN_ID_COLUMN, "0")), 16)
     trace_flags = TraceFlags(TraceFlags.SAMPLED)
     response_id = row.get("gen_ai_response_id", "")
     evaluation_results = row.get("evaluation", {})
@@ -109,6 +110,8 @@ def get_combined_data(preprocessed_data, evaluated_data, service_name):
 def run(args):
     """Entry point of model prediction script."""
     logger.info(f"Commandline args:> Service Name: {args['service_name']}")
+    if is_input_data_empty(args["preprocessed_data"]):
+        return
     provider = configure_logging(args)
     data = get_combined_data(args["preprocessed_data"], args["evaluated_data"],
                              args["service_name"])

assets/evaluation_on_cloud/environments/evaluations-built-in/context/online_eval/utils.py

@@ -4,6 +4,7 @@
 """Utility functions for the online evaluation context."""
 import os
 import re
+import pandas as pd
 
 from azure.ai.ml import MLClient
 from azure.ai.ml.identity import AzureMLOnBehalfOfCredential
@@ -116,3 +117,16 @@ def get_mlclient(
 
     logger.info(f"Creating MLClient with registry name {registry_name}")
     return MLClient(credential=credential, registry_name=registry_name)
+
+
+def is_input_data_empty(data_file_path):
+    """Check if the input data is empty."""
+    if not data_file_path:
+        logger.info("Data file path is empty. Exiting.")
+        return True
+
+    df = pd.read_json(data_file_path, lines=True)
+    if len(df) == 0:
+        logger.info("Empty data in preprocessed file. Exiting.")
+        return True
+    return False