🖨 Add Script Utilities section to the documentation (huggingface#2407)

* Add script_utils.md to the documentation

* Refactor ScriptArguments class documentation

* Refactor TrlParser class to improve code organization and readability

Author: qgallouedec

docs/source/_toctree.yml

@@ -65,6 +65,8 @@
     title: Data Utilities
   - local: text_environments
     title: Text Environments
+  - local: script_utils
+    title: Script Utilities
   title: API
 - sections:
   - local: example_overview

docs/source/script_utils.md

@@ -0,0 +1,9 @@
+# Scripts Utilities
+
+## ScriptArguments
+
+[[autodoc]] ScriptArguments
+
+## TrlParser
+
+[[autodoc]] TrlParser

trl/commands/cli_utils.py

@@ -151,26 +151,29 @@ class ChatArguments:
 
 
 class TrlParser(HfArgumentParser):
+    """
+    The TRL parser parses a list of parsers (TrainingArguments, trl.ModelConfig, etc.), creates a config
+    parsers for users that pass a valid `config` field and merge the values that are set in the config
+    with the processed parsers.
+
+    Args:
+        parsers (`List[argparse.ArgumentParser]`):
+            List of parsers.
+        ignore_extra_args (`bool`):
+            Whether to ignore extra arguments passed by the config
+            and not raise errors.
+    """
+
     def __init__(self, parsers, ignore_extra_args=False):
-        """
-        The TRL parser parses a list of parsers (TrainingArguments, trl.ModelConfig, etc.), creates a config
-        parsers for users that pass a valid `config` field and merge the values that are set in the config
-        with the processed parsers.
-
-        Args:
-            parsers (`list[argparse.ArgumentParser`]):
-                List of parsers.
-            ignore_extra_args (`bool`):
-                Whether to ignore extra arguments passed by the config
-                and not raise errors.
-        """
         super().__init__(parsers)
         self.yaml_parser = YamlConfigParser()
         self.ignore_extra_args = ignore_extra_args
 
     def post_process_dataclasses(self, dataclasses):
-        # Apply additional post-processing in case some arguments needs a special
-        # care
+        """
+        Post process dataclasses to merge the TrainingArguments with the SFTScriptArguments or DPOScriptArguments.
+        """
+
         training_args = trl_args = None
         training_args_index = None
 
@@ -192,6 +195,9 @@ def post_process_dataclasses(self, dataclasses):
         return dataclasses
 
     def parse_args_and_config(self, return_remaining_strings=False):
+        """
+        Parse the command line arguments and the config file.
+        """
         yaml_config = None
         if "--config" in sys.argv:
             config_index = sys.argv.index("--config")

trl/utils.py

@@ -21,19 +21,20 @@ class ScriptArguments:
     """
     Arguments common to all scripts.
 
-    dataset_name (`str`):
-        Dataset name.
-    dataset_train_split (`str`, *optional*, defaults to `"train"`):
-        Dataset split to use for training.
-    dataset_test_split (`str`, *optional*, defaults to `"test"`):
-        Dataset split to use for evaluation.
-    config (`str` or `None`, *optional*, defaults to `None`):
-        Path to the optional config file.
-    gradient_checkpointing_use_reentrant (`bool`, *optional*, defaults to `False`):
-        Whether to apply `use_reentrant` for gradient_checkpointing.
-    ignore_bias_buffers (`bool`, *optional*, defaults to `False`):
-        Debug argument for distributed training. Fix for DDP issues with LM bias/mask buffers - invalid scalar type,
-        inplace operation. See https://github.com/huggingface/transformers/issues/22482#issuecomment-1595790992.
+    Args:
+        dataset_name (`str`):
+            Dataset name.
+        dataset_train_split (`str`, *optional*, defaults to `"train"`):
+            Dataset split to use for training.
+        dataset_test_split (`str`, *optional*, defaults to `"test"`):
+            Dataset split to use for evaluation.
+        config (`str` or `None`, *optional*, defaults to `None`):
+            Path to the optional config file.
+        gradient_checkpointing_use_reentrant (`bool`, *optional*, defaults to `False`):
+            Whether to apply `use_reentrant` for gradient_checkpointing.
+        ignore_bias_buffers (`bool`, *optional*, defaults to `False`):
+            Debug argument for distributed training. Fix for DDP issues with LM bias/mask buffers - invalid scalar
+            type, inplace operation. See https://github.com/huggingface/transformers/issues/22482#issuecomment-1595790992.
     """
 
     dataset_name: str