Deprecate unused dataset_formatting module (#4242)

Co-authored-by: behroozazarkhalili <ermiaazarkhalili>
Co-authored-by: Quentin Gallouédec <[email protected]>

Author: behroozazarkhalili

tests/test_dataset_formatting.py

@@ -14,6 +14,7 @@
 
 from typing import Callable
 
+import pytest
 from datasets import Dataset, load_dataset
 from transformers import AutoModelForCausalLM, AutoModelForSequenceClassification, AutoTokenizer
 
@@ -23,6 +24,7 @@
 from .testing_utils import TrlTestCase
 
 
+@pytest.mark.filterwarnings("ignore::FutureWarning")
 class TestDatasetFormatting(TrlTestCase):
     def setup_method(self):
         self.llama_tokenizer = AutoTokenizer.from_pretrained("trl-internal-testing/tiny-MistralForCausalLM-0.1")

trl/extras/dataset_formatting.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 
 import logging
+import warnings
 from typing import Callable, Literal, Optional
 
 import datasets
@@ -41,7 +42,17 @@ def conversations_formatting_function(
     r"""
     return a callable function that takes in a "messages" dataset and returns a formatted dataset, based on the
     tokenizer apply chat template to the dataset along with the schema of the list of functions in the tools list.
+
+    .. deprecated:: 0.24.0
+        `conversations_formatting_function` is deprecated and will be removed in version 0.27.
+        Please use `tokenizer.apply_chat_template()` directly instead.
     """
+    warnings.warn(
+        "`conversations_formatting_function` is deprecated and will be removed in TRL 0.27. "
+        "Please use `tokenizer.apply_chat_template()` directly instead.",
+        DeprecationWarning,
+        stacklevel=2,
+    )
 
     def format_dataset(examples):
         if isinstance(examples[messages_field][0], list):
@@ -61,7 +72,17 @@ def instructions_formatting_function(tokenizer: AutoTokenizer):
     r"""
     return a callable function that takes in an "instructions" dataset and returns a formatted dataset, based on the
     tokenizer apply chat template to the dataset
+
+    .. deprecated:: 0.24.0
+        `instructions_formatting_function` is deprecated and will be removed in version 0.27.
+        Please use `tokenizer.apply_chat_template()` directly instead.
     """
+    warnings.warn(
+        "`instructions_formatting_function` is deprecated and will be removed in TRL 0.27. "
+        "Please use `tokenizer.apply_chat_template()` directly instead.",
+        DeprecationWarning,
+        stacklevel=2,
+    )
 
     def format_dataset(examples):
         if isinstance(examples["prompt"], list):
@@ -99,7 +120,18 @@ def get_formatting_func_from_dataset(
 
     Returns:
         Callable: Formatting function if the dataset format is supported else None
+
+    .. deprecated:: 0.24.0
+        `get_formatting_func_from_dataset` is deprecated and will be removed in version 0.27.
+        Please use `tokenizer.apply_chat_template()` directly instead.
     """
+    warnings.warn(
+        "`get_formatting_func_from_dataset` is deprecated and will be removed in TRL 0.27. "
+        "Please use `tokenizer.apply_chat_template()` directly instead.",
+        DeprecationWarning,
+        stacklevel=2,
+    )
+
     if isinstance(dataset, Dataset):
         if "messages" in dataset.features:
             if dataset.features["messages"] == FORMAT_MAPPING["chatml"]: