fix: crash if requested output field for inference doesn't exist in d…

…ataset (#624)

* fix: crash if requested output field for inference doesn't exist in dataset
* refactor: overwrite column is a warning, update output with dataset content
* test: both empty and entirely missing ground-truth datasets

lumigator/python/mzai/backend/backend/tests/conftest.py

@@ -94,12 +94,19 @@ def dialog_dataset():
 
 
 @pytest.fixture(scope="function")
-def dialog_no_gt_dataset():
+def dialog_empty_gt_dataset():
     filename = common_resources_dir() / "sample_data" / "dialogsum_mini_empty_gt.csv"
     with Path(filename).open("rb") as f:
         yield f
 
 
+@pytest.fixture(scope="function")
+def dialog_no_gt_dataset():
+    filename = common_resources_dir() / "sample_data" / "dialogsum_mini_no_gt.csv"
+    with Path(filename).open("rb") as f:
+        yield f
+
+
 @pytest.fixture(scope="session", autouse=True)
 def db_engine():
     """Initialize a DB engine and create tables."""

lumigator/python/mzai/backend/backend/tests/integration/api/routes/test_api_workflows.py

@@ -135,17 +135,20 @@ def test_upload_data_launch_job(
     logger.info(f"#{logs_infer_job_response_model.logs}#")
 
 
+@pytest.mark.parametrize("unnanotated_dataset", ["dialog_empty_gt_dataset", "dialog_no_gt_dataset"])
 def test_upload_data_no_gt_launch_annotation(
+    request: pytest.FixtureRequest,
     local_client: TestClient,
-    dialog_no_gt_dataset,
+    unnanotated_dataset,
     simple_eval_template,
     simple_infer_template,
     dependency_overrides_services,
 ):
+    dataset = request.getfixturevalue(unnanotated_dataset)
     create_response = local_client.post(
         "/datasets/",
         data={},
-        files={"dataset": dialog_no_gt_dataset, "format": (None, DatasetFormat.JOB.value)},
+        files={"dataset": dataset, "format": (None, DatasetFormat.JOB.value)},
     )
 
     assert create_response.status_code == 201

lumigator/python/mzai/jobs/inference/inference.py

@@ -109,10 +109,17 @@ def run_inference(config: InferenceJobConfig) -> Path:
     else:
         raise NotImplementedError("Inference pipeline not supported.")
 
-    # run inference
+    # We keep any columns that were already there (not just the original input
+    # samples, but also past predictions under another column name)
+    output.update(dataset.to_dict())
+
+    # We are trusting the user: if the dataset already had a column with the output_field
+    # they selected, we overwrite it with the values from our inference.
+
+    if config.job.output_field in dataset.column_names:
+        logger.warning(f"Overwriting {config.job.output_field}")
+
     output[config.job.output_field] = predict(dataset_iterable, model_client)
-    output["examples"] = dataset["examples"]
-    output["ground_truth"] = dataset["ground_truth"]
     output["model"] = output_model_name
 
     output_path = save_outputs(config, output)

lumigator/python/mzai/sample_data/dialogsum_mini_no_gt.csv

@@ -0,0 +1,55 @@
+examples
+"#Person1#: Hello, how are you doing today?
+#Person2#: I ' Ve been having trouble breathing lately.
+#Person1#: Have you had any type of cold lately?
+#Person2#: No, I haven ' t had a cold. I just have a heavy feeling in my chest when I try to breathe.
+#Person1#: Do you have any allergies that you know of?
+#Person2#: No, I don ' t have any allergies that I know of.
+#Person1#: Does this happen all the time or mostly when you are active?
+#Person2#: It happens a lot when I work out.
+#Person1#: I am going to send you to a pulmonary specialist who can run tests on you for asthma.
+#Person2#: Thank you for your help, doctor."
+"#Person1#: Hey Jimmy. Let's go workout later today.
+#Person2#: Sure. What time do you want to go?
+#Person1#: How about at 3:30?
+#Person2#: That sounds good. Today we work on Legs and forearm.
+#Person1#: Hey. I just played basketball earlier, so my legs are a little sore. Let's work out on arms and stomach today.
+#Person2#: I'm on a weekly schedule. You're messing everything up.
+#Person1#: C'mon. We're only switching two days. You can do legs on Friday.
+#Person2#: Aright. I'll meet you at the gym at 3:30 then."
+"#Person1#: I need to stop eating such unhealthy foods.
+#Person2#: I know what you mean. I've started eating better myself.
+#Person1#: What foods do you eat now?
+#Person2#: I tend to stick to fruits, vegetables, and chicken.
+#Person1#: Those are the only things you eat?
+#Person2#: That's basically what I eat.
+#Person1#: Why aren't you eating anything else?
+#Person2#: Well, fruits and vegetables are very healthy.
+#Person1#: And the chicken?
+#Person2#: It's really healthy to eat when you bake it.
+#Person1#: I guess that does sound a lot healthier."
+"#Person1#: Do you believe in UFOs?
+#Person2#: Of course, they are out there.
+#Person1#: But I never saw them.
+#Person2#: Are you stupid? They are called UFOs, so not everybody can see them.
+#Person1#: You mean that you can them.
+#Person2#: That's right. I can see them in my dreams.
+#Person1#: They come to the earth?
+#Person2#: No. Their task is to send the aliens here from the outer space.
+#Person1#: Aliens from the outer space? Do you talk to them? What do they look like?
+#Person2#: OK, OK, one by one, please! They look like robots, but they can speak. Their mission is to make friends with human beings.
+#Person1#: That means that you talk to them? In which language?
+#Person2#: Of course in English, they learn English on Mars too.
+#Person1#: Wow. Sounds fantastic!"
+"#Person1#: Did you go to school today?
+#Person2#: Of course. Did you?
+#Person1#: I didn't want to, so I didn't.
+#Person2#: That's sad, but have you gone to the movies recently?
+#Person1#: That's a switch.
+#Person2#: I'm serious, have you?
+#Person1#: No, I haven't. Why?
+#Person2#: I really want to go to the movies this weekend.
+#Person1#: So go then.
+#Person2#: I really don't want to go by myself.
+#Person1#: Well anyway, do you plan on going to school tomorrow?
+#Person2#: No, I think I'm going to go to the movies."