MLFlow DB export now only includes the last metric values

drastically decreases result query times

evaluate.sh

@@ -20,13 +20,15 @@ docker exec -it $USER $CUDA_ENV --workdir /app/src $CONTAINER_ID python3 ./usgoc
 	# | grep --line-buffered -vE \
 	# "BaseCollectiveExecutor::StartAbort|IteratorGetNext|Shape/|Shape_[0-9]+/"
 
+succ=$?
+
 if [ ! -z "$NO_EXPORT" ]; then
 	echo "No MLFLow DB export."
-elif [ $? -eq 0 ]; then
+elif [ $succ -eq 0 ]; then
 	echo "Pruning MLFLow db..."
 	EXPERIMENT_NAME=${EXPERIMENT_NAME:-$(cat EXPERIMENT_NAME)}
 	docker exec -it $USER --workdir /app $CONTAINER_ID mlflow gc --backend-store-uri file:./mlruns
-	docker exec -it $USER -e EXPERIMENT_NAME=$EXPERIMENT_NAME --workdir /app $CONTAINER_ID mlflow_migrate/sqlite_migrate.sh
+	docker exec -it $USER -e EXPERIMENT_NAME=$EXPERIMENT_NAME -e ONLY_LAST_METRIC=1 --workdir /app $CONTAINER_ID mlflow_migrate/sqlite_migrate.sh
 else
 	echo "Evaluation failed."
 	exit 1

src/usgoc/run_evaluation.py

@@ -43,11 +43,16 @@
 @click.option("--yes", "-y", is_flag=True, default=False)
 @click.option("--export", "-e", is_flag=True, default=False)
 @click.option("--confusion-matrices", is_flag=True, default=False)
+@click.option("--nop", is_flag=True, default=False)
 def evaluate(
   model, convert_mode, limit_id, fold=None, repeat=None,
   override=False, dry=False,
   tuner_convert_mode=None, tuner_limit_id=None,
-  suffix="", yes=False, export=False, confusion_matrices=False):
+  suffix="", yes=False, export=False, confusion_matrices=False, nop=False):
+  if nop:
+    print("NO OP: No evaluation done.")
+    return
+
   if export:
     print("Starting export of winning models.")
     f = ee.export_best