Remove post-collation truncation from SFT

tests/test_sft_trainer.py

@@ -1003,17 +1003,16 @@ def test_skip_prepare_dataset_passes_truncation_to_text_collator(self):
         assert trainer.data_collator.max_length == 16
         assert trainer.data_collator.truncation_mode == "keep_end"
 
-    def test_skip_prepare_dataset_with_padding_free_and_max_length_raises(self):
+    def test_padding_free_without_packing_and_max_length_raises(self):
         dataset = load_dataset("trl-internal-testing/zen", "standard_language_modeling", split="train[:2]")
         training_args = SFTConfig(
             output_dir=self.tmp_dir,
             max_length=16,
             padding_free=True,
-            dataset_kwargs={"skip_prepare_dataset": True},
             report_to="none",
         )
 
-        with pytest.raises(ValueError, match="must be enforced during dataset preparation or packing"):
+        with pytest.raises(ValueError, match="max_length` is not enforced"):
             SFTTrainer(
                 model="trl-internal-testing/tiny-Qwen2ForCausalLM-2.5", args=training_args, train_dataset=dataset
             )

trl/trainer/sft_trainer.py

@@ -51,7 +51,6 @@
     maybe_convert_to_chatml,
     pack_dataset,
     prepare_multimodal_messages,
-    truncate_dataset,
 )
 from ..models import get_act_offloading_ctx_manager
 from .base_trainer import _BaseTrainer
@@ -572,6 +571,8 @@ class SFTTrainer(_BaseTrainer):
             Function to use to form a batch from a list of elements of the processed `train_dataset` or `eval_dataset`.
             Will default to [`~trainer.sft_trainer.DataCollatorForLanguageModeling`] if the model is a language model
             and [`~trainer.sft_trainer.DataCollatorForVisionLanguageModeling`] if the model is a vision-language model.
+            Custom collators must truncate sequences before padding; the trainer does not apply post-collation
+            truncation.
         train_dataset ([`~datasets.Dataset`] or [`~datasets.IterableDataset`]):
             Dataset to use for training. This trainer supports both [language modeling](#language-modeling) type and
             [prompt-completion](#prompt-completion) type. The format of the samples can be either:
@@ -913,19 +914,19 @@ def __init__(
             )
 
         # Dataset
+        if self.padding_free and not args.packing and args.max_length is not None and not self._is_vision_dataset:
+            raise ValueError(
+                "When `padding_free=True` without packing, `max_length` is not enforced. Either enable packing "
+                "(e.g., `packing=True, packing_strategy='bfd'`), provide already truncated inputs, or set "
+                "`max_length=None`."
+            )
         # Skip dataset preparation if `skip_prepare_dataset=True` in `dataset_kwargs`, or if it's a VLM, where
         # preprocessing (e.g., image-to-pixel conversion) is too costly and done on the fly instead.
         skip_prepare_dataset = (
             args.dataset_kwargs is not None
             and args.dataset_kwargs.get("skip_prepare_dataset", False)
             or self._is_vision_dataset
         )
-        if skip_prepare_dataset and self.padding_free and args.max_length is not None and not self._is_vision_dataset:
-            raise ValueError(
-                "When `padding_free=True`, `max_length` must be enforced during dataset preparation or packing, not in "
-                "the collator. Disable `skip_prepare_dataset`, provide already packed/truncated inputs, or set "
-                "`max_length=None`."
-            )
         if not skip_prepare_dataset:
             if self.completion_only_loss and formatting_func:
                 raise ValueError(
@@ -1188,7 +1189,7 @@ def tokenize_fn(example, processing_class, dataset_text_field, assistant_only_lo
                     **map_kwargs,
                 )
 
-            # Pack or truncate
+            # Pack
             if packing:
                 if args.max_length is None:
                     raise ValueError("When packing is enabled, `max_length` can't be `None`.")
@@ -1210,12 +1211,6 @@ def tokenize_fn(example, processing_class, dataset_text_field, assistant_only_lo
 
                 # Packing adds new column "seq_lengths" needed for document aware FlashAttention
                 dataset = pack_dataset(dataset, args.max_length, args.packing_strategy, map_kwargs)
-            elif args.max_length is not None:
-                if isinstance(dataset, Dataset):  # `IterableDataset.map` does not support `desc`
-                    map_kwargs["desc"] = f"Truncating {dataset_name} dataset"
-                dataset = truncate_dataset(
-                    dataset, args.max_length, truncation_mode=args.truncation_mode, map_kwargs=map_kwargs
-                )
             # For Liger kernel, ensure only the essential columns
             if args.use_liger_kernel:
                 collator_expected_keys = {"input_ids", "seq_lengths", "completion_mask", "assistant_masks"}