added changes for MOE in export and download

backend/dataset/migrations/0051_auto_20240725_0900.py

@@ -0,0 +1,26 @@
+# Generated by Django 3.2.14 on 2024-07-25 09:00
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ("dataset", "0050_alter_interaction_no_of_turns"),
+    ]
+
+    operations = [
+        migrations.RemoveField(
+            model_name="promptanswer",
+            name="eval_output_likert_score",
+        ),
+        migrations.AddField(
+            model_name="promptanswer",
+            name="prompt_output_pair_id",
+            field=models.CharField(
+                help_text="prompt_output_pair_id",
+                max_length=16,
+                null=True,
+                verbose_name="prompt_output_pair_id",
+            ),
+        ),
+    ]

backend/dataset/models.py

@@ -788,12 +788,6 @@ class PromptAnswer(DatasetBase):
     language = models.CharField(
         verbose_name="language", choices=LANG_CHOICES, max_length=15
     )
-    eval_output_likert_score = models.IntegerField(
-        verbose_name="evaluation_prompt_response_rating",
-        null=True,
-        blank=True,
-        help_text=("Rating of the prompt response"),
-    )
     eval_form_output_json = models.JSONField(
         verbose_name="evaluation_form_output",
         null=True,
@@ -806,6 +800,12 @@ class PromptAnswer(DatasetBase):
         blank=True,
         help_text=("Time taken to complete the prompt response"),
     )
+    prompt_output_pair_id = models.CharField(
+        verbose_name="prompt_output_pair_id",
+        max_length=16,
+        help_text=("prompt_output_pair_id"),
+        null=True,
+    )
 
     def __str__(self):
         return str(self.id)

backend/projects/project_registry.yaml

@@ -37,7 +37,7 @@ ModelOutputEvaluation:
             id: interaction_id
           annotations:
             - eval_form_output_json
-            - eval_output_likert_score
+            - prompt_output_pair_id
             - eval_time_taken
             - model
             - prompt

backend/projects/tasks.py

@@ -685,7 +685,7 @@ def export_project_new_record(
             )
 
         elif project_type == "ModelInteractionEvaluation":
-            item_data_list = get_attributes_for_ModelInteractionEvaluation(task, True)
+            item_data_list = get_attributes_for_ModelInteractionEvaluation(task)
             for item in range(len(item_data_list)):
                 data_item = dataset_model()
                 data_item.instance_id = export_dataset_instance

backend/projects/utils.py

@@ -228,23 +228,8 @@ def ocr_word_count(annotation_result):
     return word_count
 
 
-# function to obtain the correct annotation object
-def get_correct_annotation_obj(task):
-    annotation = Annotation.objects.filter(task=task)
-    correct_ann_obj = annotation[0]
-    if len(annotation) == 2:
-        for ann in annotation:
-            if ann.annotation_type == REVIEWER_ANNOTATION:
-                correct_ann_obj = ann
-    elif len(annotation) == 3:
-        for ann in annotation:
-            if ann.annotation_type == SUPER_CHECKER_ANNOTATION:
-                correct_ann_obj = ann
-    return correct_ann_obj
-
-
 def get_attributes_for_IDC(project, task):
-    correct_ann_obj = get_correct_annotation_obj(task)
+    correct_ann_obj = task.correct_annotation
     result_dict = {
         "interactions_json": correct_ann_obj.result,
         "language": project.tgt_language,
@@ -264,55 +249,50 @@ def get_prompt_output_by_id(prompt_output_pair_id, task_data_dict):
     return None, None
 
 
-def get_attributes_for_ModelInteractionEvaluation(task, correction_annotation_present):
+def get_attributes_for_ModelInteractionEvaluation(task):
     res = []
-    if correction_annotation_present:
-        correct_ann_obj = get_correct_annotation_obj(task)
+    if task.correct_annotation:
+        correct_ann_obj = task.correct_annotation
         annotation_result_json = correct_ann_obj.result
         interaction = Interaction.objects.get(id=task.data["interaction_id"])
-        try:
-            prompt_output_pair = get_prompt_output_by_id(
-                annotation_result_json["prompt_output_pair_id"], task.data
-            )
-        except Exception as e:
-            prompt_output_pair = ["", ""]
     else:
-        annotation_result_json = task["annotations"][0]["result"]
+        annotation_result_json = Annotation.objects.filter(task=task)[0].result
         annotation_result_json = (
             json.loads(annotation_result_json)
             if isinstance(annotation_result_json, str)
             else annotation_result_json
         )
-        interaction = Interaction.objects.get(id=task["data"]["interaction_id"])
+        interaction = Interaction.objects.get(id=task.data["interaction_id"])
+    for a in annotation_result_json:
         try:
-            prompt_output_pair = [
-                annotation_result_json["prompt"],
-                annotation_result_json["output"],
-            ]
+            prompt_output_pair = get_prompt_output_by_id(
+                a["prompt_output_pair_id"], task.data
+            )
         except Exception as e:
-            prompt_output_pair = ["", ""]
-
-    model = interaction.model
-    language = interaction.language
-
-    temp_attributes_obj = {
-        "interaction_id": interaction,
-        "model": model,
-        "language": language,
-        "prompt": prompt_output_pair[0],
-        "output": prompt_output_pair[1],
-    }
-    if "questions_response" in annotation_result_json:
-        temp_attributes_obj["eval_form_output_json"] = annotation_result_json[
-            "questions_response"
-        ]
-    if "rating" in annotation_result_json:
-        temp_attributes_obj["eval_output_likert_score"] = annotation_result_json[
-            "rating"
-        ]
-    if "time_taken" in annotation_result_json:
-        temp_attributes_obj["eval_time_taken"] = annotation_result_json["time_taken"]
-    res.append(temp_attributes_obj)
+            try:
+                prompt_output_pair = [
+                    annotation_result_json["prompt"],
+                    annotation_result_json["output"],
+                ]
+            except Exception as e:
+                prompt_output_pair = ["", ""]
+        model = interaction.model
+        language = interaction.language
+
+        temp_attributes_obj = {
+            "interaction_id": interaction,
+            "model": model,
+            "language": language,
+            "prompt": prompt_output_pair[0],
+            "output": prompt_output_pair[1],
+        }
+        if "prompt_output_pair_id" in a:
+            temp_attributes_obj["prompt_output_pair_id"] = a["prompt_output_pair_id"]
+        if "questions_response" in a:
+            temp_attributes_obj["eval_form_output_json"] = a["questions_response"]
+        if "time_taken" in a:
+            temp_attributes_obj["eval_time_taken"] = a["time_taken"]
+        res.append(temp_attributes_obj)
 
     return res
 

backend/projects/views.py

@@ -3968,15 +3968,6 @@ def download(self, request, pk=None, *args, **kwargs):
                     )
                     task["data"]["interactions_json"] = annotation_result
                     del task["annotations"]
-            elif dataset_type == "Interaction":
-                for task in tasks_list:
-                    item_data_list = get_attributes_for_ModelInteractionEvaluation(
-                        task, False
-                    )
-                    for it in item_data_list:
-                        for key, value in it.items():
-                            task["data"][key] = value
-                    del task["annotations"]
             return DataExport.generate_export_file(project, tasks_list, export_type)
         except Project.DoesNotExist:
             ret_dict = {"message": "Project does not exist!"}