feat(data): add Hermes function-calling dataset preset

Register NousResearch/hermes-function-calling-v1 (func_calling config) in both
dataset registries: the offline preprocessing DATASET_CONFIGS (configs.py) and
the on-policy response-regeneration script's DATASET_CONFIGS.

Hermes func_calling is a multi-turn function-calling SFT dataset (Apache-2.0) in
ShareGPT from/value form, so it needs no normalizer (like the existing sharegpt
preset). Tool calls/results are text-encoded (<tool_call>/<tool_response>)
inline; the parallel `tools` column carries the JSON schemas as a JSON string.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LRRGNt2T9ktWyFRboNUZNU
Signed-off-by: Ranran Haoran Zhang <ranzhang@redhat.com>

Author: WindChimeRan

scripts/response_regeneration/script.py

@@ -35,6 +35,12 @@
         "default_split": "train",
         "subset": "main",
     },
+    "hermes-fc": {
+        "id": "NousResearch/hermes-function-calling-v1",
+        "prompt_field": "conversations",
+        "default_split": "train",
+        "subset": "func_calling",
+    },
 }
 
 

src/speculators/data_generation/configs.py

@@ -146,4 +146,11 @@ def _normalize_sharegpt4v_coco(example: dict) -> dict:
         hf_path="mlabonne/open-perfectblend",
         split="train",
     ),
+    # Multi-turn function-calling SFT
+    "hermes-fc": DatasetConfig(
+        name="hermes-fc",
+        hf_path="NousResearch/hermes-function-calling-v1",
+        subset="func_calling",
+        split="train",
+    ),
 }