use max of 32 dataset processes if not explicit (#2403)

* use max of 32 dataset processes if not explicit

* change alternate min val for consistency

src/axolotl/core/datasets/chat.py

@@ -43,7 +43,7 @@ class TokenizedChatDataset(Dataset):
         process_or_cpu_count: int = (
             process_count or os.cpu_count()  # type: ignore[assignment]
         )
-        num_proc = min(64, process_or_cpu_count)
+        num_proc = min(32, process_or_cpu_count)
         features = data.features.keys()
         tokenized_data = data.map(
             map_fn,

src/axolotl/utils/config/models/input/v0_4_1/__init__.py

@@ -728,7 +728,7 @@ class AxolotlInputConfig(
         default=None,
         json_schema_extra={"description": "streaming dataset to use for pretraining"},
     )
-    dataset_processes: Optional[int] = Field(default=os.cpu_count())
+    dataset_processes: Optional[int] = Field(default=min(32, os.cpu_count()))  # type: ignore[type-var]
     dataset_exact_deduplication: Optional[bool] = None
     dataset_keep_in_memory: Optional[bool] = None
     dataloader_pin_memory: Optional[bool] = None