feat(components): use rlhf_preprocessor to replace the current value_exists call in rlhf

PiperOrigin-RevId: 627630447

Author: Googler

components/google-cloud/google_cloud_pipeline_components/_implementation/llm/generated/refined_image_versions.py

@@ -17,4 +17,4 @@
 DO NOT EDIT - This file is generated, manual changes will be overridden.
 """
 
-IMAGE_TAG = '20240417_0507_RC00'
+IMAGE_TAG = '20240423_1336_RC00'

components/google-cloud/google_cloud_pipeline_components/preview/llm/rlhf/component.py

@@ -191,11 +191,10 @@ def rlhf_pipeline(
       encryption_spec_key_name=encryption_spec_key_name,
   ).set_display_name('Reinforcement Learning')
 
-  has_inference_dataset = function_based.value_exists(
-      value=eval_dataset
-  ).set_display_name('Resolve Inference Dataset')
+  has_inference_dataset = preprocess_metadata.outputs['has_inference_dataset']
+
   with kfp.dsl.Condition(
-      has_inference_dataset.output == True,  # pylint: disable=singleton-comparison
+      has_inference_dataset == True,  # pylint: disable=singleton-comparison
       name='Perform Inference',
   ):
     has_model_checkpoint = function_based.value_exists(