reduce benchmark profiling to generating the dataset only. Not inferr…

…ing (that is dome mocking anyhow) and not evaluating (of the mocked results)

Signed-off-by: dafnapension <[email protected]>

performance/bluebench_profiler.py

@@ -15,7 +15,6 @@
     TextGenerationInferenceOutput,
 )
 from unitxt.logging_utils import get_logger
-from unitxt.schema import UNITXT_DATASET_SCHEMA, loads_instance
 from unitxt.settings_utils import get_settings
 
 logger = get_logger()
@@ -72,12 +71,12 @@ def profiler_generate_benchmark_dataset(
         ):
             stream = benchmark_recipe()[split]
 
-            dataset = stream.to_dataset(
-                features=UNITXT_DATASET_SCHEMA, disable_cache=False
-            ).with_transform(loads_instance)
+            # dataset = stream.to_dataset(
+            #     features=UNITXT_DATASET_SCHEMA, disable_cache=False
+            # ).with_transform(loads_instance)
 
             # to charge here for the time of generating all instances
-            return list(dataset)
+            return list(stream)
 
     def profiler_instantiate_model(self) -> InferenceEngine:
         return CrossProviderInferenceEngine(
@@ -102,14 +101,14 @@ def profiler_do_the_profiling(self, dataset_query: str, split: str, **kwargs):
             benchmark_recipe=benchmark_recipe, split=split, **kwargs
         )
 
-        model = self.profiler_instantiate_model()
+        # model = self.profiler_instantiate_model()
 
-        predictions = self.profiler_infer_predictions(model=model, dataset=dataset)
+        # predictions = self.profiler_infer_predictions(model=model, dataset=dataset)
 
-        evaluation_result = self.profiler_evaluate_predictions(
-            predictions=predictions, dataset=dataset
-        )
-        logger.critical(f"length of evaluation_result: {len(evaluation_result)}")
+        # evaluation_result = self.profiler_evaluate_predictions(
+        #     predictions=predictions, dataset=dataset
+        # )
+        logger.critical(f"length of evaluation_result: {len(dataset)}")
 
 
 dataset_query = "benchmarks.bluebench[loader_limit=30,max_samples_per_subset=30]"