fix: catch ValueError in parquet fallback for HF datasets compat

HF datasets raises ValueError (not just TypeError) when it encounters
unknown feature types in embedded parquet metadata. Catch both so the
PyArrow fallback triggers correctly.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Ye Yu <yeyu@nvidia.com>

Author: yeyu-nvidia

examples/specdec_bench/specdec_bench/datasets/speed.py

@@ -718,7 +718,7 @@ def _load_dataset(self, config_name_or_dataset_path: config_type | str) -> "Data
                 data_files = {"test": [str(config_name_or_dataset_path_path)]}
             try:
                 dataset = load_dataset("parquet", data_files=data_files, split="test")
-            except TypeError:
+            except (TypeError, ValueError):
                 # Fallback: parquet metadata may be incompatible with the installed
                 # ``datasets`` version.  Read via PyArrow and convert directly.
                 import pyarrow