Revert frozen training arguments (#25903)

* Revert frozen training arguments

* TODO

examples/pytorch/image-pretraining/run_mae.py

@@ -163,15 +163,6 @@ class CustomTrainingArguments(TrainingArguments):
         default=1e-3, metadata={"help": "Base learning rate: absolute_lr = base_lr * total_batch_size / 256."}
     )
 
-    def __post_init__(self):
-        # Compute absolute learning rate while args are mutable
-        super().__post_init__()
-        if self.base_learning_rate is not None:
-            total_train_batch_size = self.train_batch_size * self.gradient_accumulation_steps * self.world_size
-            delattr(self, "_frozen")
-            self.learning_rate = self.base_learning_rate * total_train_batch_size / 256
-            setattr(self, "_frozen", True)
-
 
 def collate_fn(examples):
     pixel_values = torch.stack([example["pixel_values"] for example in examples])
@@ -362,6 +353,13 @@ def main():
         # Set the validation transforms
         ds["validation"].set_transform(preprocess_images)
 
+    # Compute absolute learning rate
+    total_train_batch_size = (
+        training_args.train_batch_size * training_args.gradient_accumulation_steps * training_args.world_size
+    )
+    if training_args.base_learning_rate is not None:
+        training_args.learning_rate = training_args.base_learning_rate * total_train_batch_size / 256
+
     # Initialize our trainer
     trainer = Trainer(
         model=model,

examples/pytorch/summarization/run_summarization.py

@@ -18,7 +18,6 @@ Fine-tuning the library models for sequence to sequence.
 """
 # You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments.
 
-import dataclasses
 import logging
 import os
 import sys
@@ -675,10 +674,14 @@ def main():
         return result
 
     # Override the decoding parameters of Seq2SeqTrainer
-    if training_args.generation_max_length is None:
+    training_args.generation_max_length = (
-        training_args = dataclasses.replace(training_args, generation_max_length=data_args.val_max_target_length)
+        training_args.generation_max_length
-    if training_args.generation_num_beams is None:
+        if training_args.generation_max_length is not None
-        training_args = dataclasses.replace(training_args, generation_num_beams=data_args.num_beams)
+        else data_args.val_max_target_length
+    )
+    training_args.generation_num_beams = (
+        data_args.num_beams if data_args.num_beams is not None else training_args.generation_num_beams
+    )
 
     # Initialize our Trainer
     trainer = Seq2SeqTrainer(

examples/research_projects/mlm_wwm/run_mlm_wwm.py

@@ -21,7 +21,6 @@ https://huggingface.co/models?filter=fill-mask
 """
 # You can also adapt this script on your own masked language modeling task. Pointers for this are left as comments.
 
-import dataclasses
 import json
 import logging
 import math
@@ -367,7 +366,7 @@ def main():
     # If we have ref files, need to avoid it removed by trainer
     has_ref = data_args.train_ref_file or data_args.validation_ref_file
     if has_ref:
-        training_args = dataclasses.replace(training_args, remove_unused_columns=False)
+        training_args.remove_unused_columns = False
 
     # Data collator
     # This one will take care of randomly masking the tokens.

examples/tensorflow/language-modeling/run_clm.py

@@ -259,6 +259,7 @@ def main():
             assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, json or txt file."
 
     if training_args.output_dir is not None:
+        training_args.output_dir = Path(training_args.output_dir)
         os.makedirs(training_args.output_dir, exist_ok=True)
     # endregion
 
@@ -266,8 +267,8 @@ def main():
     # Detecting last checkpoint.
     checkpoint = None
     if len(os.listdir(training_args.output_dir)) > 0 and not training_args.overwrite_output_dir:
-        config_path = Path(training_args.output_dir) / CONFIG_NAME
+        config_path = training_args.output_dir / CONFIG_NAME
-        weights_path = Path(training_args.output_dir) / TF2_WEIGHTS_NAME
+        weights_path = training_args.output_dir / TF2_WEIGHTS_NAME
         if config_path.is_file() and weights_path.is_file():
             checkpoint = training_args.output_dir
             logger.info(

examples/tensorflow/language-modeling/run_mlm.py

@@ -265,6 +265,7 @@ def main():
             assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, json or txt file."
 
     if training_args.output_dir is not None:
+        training_args.output_dir = Path(training_args.output_dir)
         os.makedirs(training_args.output_dir, exist_ok=True)
 
     if isinstance(training_args.strategy, tf.distribute.TPUStrategy) and not data_args.pad_to_max_length:
@@ -276,8 +277,8 @@ def main():
     # Detecting last checkpoint.
     checkpoint = None
     if len(os.listdir(training_args.output_dir)) > 0 and not training_args.overwrite_output_dir:
-        config_path = Path(training_args.output_dir) / CONFIG_NAME
+        config_path = training_args.output_dir / CONFIG_NAME
-        weights_path = Path(training_args.output_dir) / TF2_WEIGHTS_NAME
+        weights_path = training_args.output_dir / TF2_WEIGHTS_NAME
         if config_path.is_file() and weights_path.is_file():
             checkpoint = training_args.output_dir
             logger.warning(

src/transformers/trainer.py

@@ -1172,8 +1172,6 @@ class Trainer:
         elif self.hp_search_backend == HPSearchBackend.WANDB:
             params = trial
 
-        # Unfreeze args for hyperparameter search
-        delattr(self.args, "_frozen")
         for key, value in params.items():
             if not hasattr(self.args, key):
                 logger.warning(
@@ -1205,8 +1203,6 @@ class Trainer:
             self.args.hf_deepspeed_config.trainer_config_process(self.args)
             self.args.deepspeed_plugin = DeepSpeedPlugin(hf_ds_config=self.args.hf_deepspeed_config)
 
-        # Re-freeze them
-        setattr(self.args, "_frozen", True)
         self.create_accelerator_and_postprocess()
 
     def _report_to_hp_search(self, trial: Union["optuna.Trial", Dict[str, Any]], step: int, metrics: Dict[str, float]):

src/transformers/training_args.py

@@ -18,7 +18,7 @@ import json
 import math
 import os
 import warnings
-from dataclasses import FrozenInstanceError, asdict, dataclass, field, fields
+from dataclasses import asdict, dataclass, field, fields
 from datetime import timedelta
 from enum import Enum
 from pathlib import Path
@@ -154,6 +154,7 @@ class OptimizerNames(ExplicitEnum):
     PAGED_LION_8BIT = "paged_lion_8bit"
 
 
+# TODO: `TrainingArguments` users rely on it being fully mutable. In the future see if we can narrow this to a few keys: https://github.com/huggingface/transformers/pull/25903
 @dataclass
 class TrainingArguments:
     """
@@ -1707,16 +1708,6 @@ class TrainingArguments:
                 FutureWarning,
             )
 
-        # Finally set the `TrainingArguments` to be immutable
-        self._frozen = True
-
-    def __setattr__(self, name, value):
-        # Once fully through the `__post_init__`, `TrainingArguments` are immutable
-        if not name.startswith("_") and getattr(self, "_frozen", False):
-            raise FrozenInstanceError(f"cannot assign to field {name}")
-        else:
-            super().__setattr__(name, value)
-
     def __str__(self):
         self_as_dict = asdict(self)
 

tests/trainer/test_trainer.py

@@ -139,9 +139,9 @@ class RegressionTrainingArguments(TrainingArguments):
     b: float = 0.0
 
     def __post_init__(self):
+        super().__post_init__()
         # save resources not dealing with reporting (also avoids the warning when it's not set)
         self.report_to = []
-        super().__post_init__()
 
 
 class RepeatDataset:
@@ -529,8 +529,7 @@ class TrainerIntegrationPrerunTest(TestCasePlus, TrainerIntegrationCommon):
         self.check_trained_model(trainer.model)
 
         # Re-training should restart from scratch, thus lead the same results and new seed should be used.
-        args = TrainingArguments("./regression", learning_rate=0.1, seed=314)
+        trainer.args.seed = 314
-        trainer = Trainer(args=args, train_dataset=train_dataset, model_init=lambda: RegressionModel())
         trainer.train()
         self.check_trained_model(trainer.model, alternate_seed=True)
 

tests/trainer/test_trainer_distributed.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import dataclasses
 from typing import Dict
 
 import numpy as np
@@ -206,14 +205,7 @@ if __name__ == "__main__":
             logger.error(p.metrics)
             exit(1)
 
-        training_args = dataclasses.replace(training_args, eval_accumulation_steps=2)
+        trainer.args.eval_accumulation_steps = 2
-        trainer = Trainer(
-            model=DummyModel(),
-            args=training_args,
-            data_collator=DummyDataCollator(),
-            eval_dataset=dataset,
-            compute_metrics=compute_metrics,
-        )
 
         metrics = trainer.evaluate()
         logger.info(metrics)
@@ -227,22 +219,15 @@ if __name__ == "__main__":
             logger.error(p.metrics)
             exit(1)
 
-        training_args = dataclasses.replace(training_args, eval_accumulation_steps=None)
+        trainer.args.eval_accumulation_steps = None
-        trainer = Trainer(
-            model=DummyModel(),
-            args=training_args,
-            data_collator=DummyDataCollator(),
-            eval_dataset=dataset,
-            compute_metrics=compute_metrics,
-        )
 
     # Check that `dispatch_batches=False` will work on a finite iterable dataset
 
     train_dataset = FiniteIterableDataset(label_names=["labels", "extra"], length=1)
 
     model = RegressionModel()
-    training_args = dataclasses.replace(
+    training_args.per_device_train_batch_size = 1
-        training_args, per_device_train_batch_size=1, max_steps=1, dispatch_batches=False
+    training_args.max_steps = 1
-    )
+    training_args.dispatch_batches = False
     trainer = Trainer(model, training_args, train_dataset=train_dataset)
     trainer.train()