Fixes for request formatter not being overwritten correctly from scenarios with CLI and expand out dataloader info printed to console

Signed-off-by: Mark Kurtz <mark.kurtz@neuralmagic.com>

Author: markurtz

src/guidellm/__main__.py

@@ -389,11 +389,12 @@ def run(**kwargs):
     # Handle remapping for request params
     request_type = kwargs.pop("request_type", None)
     request_formatter_kwargs = kwargs.pop("request_formatter_kwargs", None)
-    kwargs["data_request_formatter"] = (
-        request_type
-        if not request_formatter_kwargs
-        else {"request_type": request_type, **request_formatter_kwargs}
-    )
+    if request_type is not None:
+        kwargs["data_request_formatter"] = (
+            request_type
+            if not request_formatter_kwargs
+            else {"request_type": request_type, **request_formatter_kwargs}
+        )
 
     # Handle output path remapping
     if (output_path := kwargs.pop("output_path", None)) is not None:

src/guidellm/benchmark/entrypoints.py

@@ -303,7 +303,7 @@ async def resolve_request_loader(
             title=(
                 f"Request loader initialized with "
                 f"{data_samples if data_samples > 0 else 'inf'} "
-                f"unique requests from {data}"
+                "unique requests"
             ),
             details=InfoMixin.extract_from_obj(request_loader),
             status_level="success",

src/guidellm/data/loaders.py

@@ -13,6 +13,7 @@
 from guidellm.data.deserializers import DatasetDeserializerFactory
 from guidellm.data.preprocessors import DataDependentPreprocessor, DatasetPreprocessor
 from guidellm.logger import logger
+from guidellm.utils import InfoMixin
 
 __all__ = ["DataLoader", "DatasetsIterator"]
 
@@ -117,7 +118,7 @@ def generator(
             )
 
 
-class DataLoader(PyTorchDataLoader[DataT]):
+class DataLoader(PyTorchDataLoader[DataT], InfoMixin):
     def __init__(
         self,
         data: list[Any],
@@ -139,6 +140,18 @@ def __init__(
             preprocessors=preprocessors,
             random_seed=random_seed,
         )
+        self._info: dict[str, Any] = {
+            "data": str(data),
+            "data_args": str(data_args),
+            "data_samples": data_samples,
+            "preprocessors": [
+                preprocessor.__class__.__name__ for preprocessor in preprocessors
+            ],
+            "collator": collator.__class__.__name__,
+            "sampler": str(sampler),
+            "num_workers": num_workers,
+            "random_seed": random_seed,
+        }
 
         super().__init__(
             dataset=iterator,
@@ -149,3 +162,7 @@ def __init__(
             num_workers=num_workers or 0,
             **kwargs,
         )
+
+    @property
+    def info(self) -> dict[str, Any]:
+        return self._info