Add basic prefix column

Signed-off-by: Samuel Monson <[email protected]>

Author: sjmonson

src/guidellm/data/objects.py

@@ -31,6 +31,7 @@
 GenerativeDatasetColumnType = Literal[
     "prompt_tokens_count_column",
     "output_tokens_count_column",
+    "prefix_column",
     "text_column",
     "image_column",
     "video_column",
@@ -195,6 +196,7 @@ class GenerativeDatasetArgs(StandardBaseDict):
     split: str | None = None
     prompt_tokens_count_column: str | None = None
     output_tokens_count_column: str | None = None
+    prefix_column: str | None = None
     text_column: str | list[str] | None = None
     image_column: str | list[str] | None = None
     video_column: str | list[str] | None = None

src/guidellm/data/utils.py

@@ -80,6 +80,11 @@
 DEFAULT_COLUMN_NAMES: dict[str, list[str]] = {
     "prompt_tokens_count": ["prompt_tokens_count", "input_tokens_count"],
     "output_tokens_count": ["output_tokens_count", "completion_tokens_count"],
+    "prefix_column": [
+        "system_prompt",
+        "system",
+        "prefix",
+    ],
     "text_column": [
         "prompt",
         "instruction",