fix(components): Pass moddel name to eval_runner to process batch pre…

…diction's output as per the output schema of model used Signed-off-by: Googler <[email protected]> PiperOrigin-RevId: 665977093
kubeflow · Aug 21, 2024 · 0d3e79a · 0d3e79a
1 parent 43cdc20
commit 0d3e79a
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 0 deletions.
diff --git a/...le_cloud_pipeline_components/_implementation/model_evaluation/llm_evaluation/component.py b/...le_cloud_pipeline_components/_implementation/model_evaluation/llm_evaluation/component.py
@@ -32,6 +32,7 @@ def model_evaluation_text_generation(
     row_based_metrics: Output[Metrics],
     project: str,
     location: str,
+    model_name: str,
     evaluation_task: str = 'text-generation',
     target_field_name: str = 'instance.output_text',
     prediction_field_name: str = 'predictions.content',
@@ -55,6 +56,7 @@ def model_evaluation_text_generation(
   Args:
     project: The GCP project that runs the pipeline component.
     location: The GCP region that runs the pipeline component.
+    model_name: The name of the model to be evaluated.
     evaluation_task: The task that the large language model will be evaluated
       on. The evaluation component computes a set of metrics relevant to that
       specific task. Currently supported tasks are: `summarization`,
@@ -124,6 +126,7 @@ def model_evaluation_text_generation(
           machine_type=machine_type,
           image_uri=version.LLM_EVAL_IMAGE_TAG,
           args=[
+              f'--model_name={model_name}',
               f'--evaluation_task={evaluation_task}',
               f'--target_field_name={target_field_name}',
               f'--prediction_field_name={prediction_field_name}',

diff --git a/..._cloud_pipeline_components/v1/model_evaluation/evaluation_llm_text_generation_pipeline.py b/..._cloud_pipeline_components/v1/model_evaluation/evaluation_llm_text_generation_pipeline.py
@@ -151,6 +151,7 @@ def evaluation_llm_text_generation_pipeline(  # pylint: disable=dangerous-defaul
   eval_task = LLMEvaluationTextGenerationOp(
       project=project,
       location=location,
+      model_name=model_name,
       evaluation_task=evaluation_task,
       target_field_name=target_field_name,
       predictions_format=batch_predict_predictions_format,