⚰️ Deprecate liger-kernel (#2949)

* Deprecate liger

* remove import

* oops, shouldn't be here

* Fix other deprecations

* remove liger from gkd for now

* remove liger for teacher

---------

Co-authored-by: Kashif Rasul <[email protected]>

docs/source/sft_trainer.md

@@ -604,17 +604,17 @@ With great memory reduction, you can potentially turn off cpu_offloading or grad
 | ![Speed up](https://raw.githubusercontent.com/linkedin/Liger-Kernel/main/docs/images/e2e-tps.png) | ![Memory](https://raw.githubusercontent.com/linkedin/Liger-Kernel/main/docs/images/e2e-memory.png) |
 
 
-1. To use Liger-Kernel in `SFTTrainer`, first install by 
+1. To use Liger-Kernel in [`SFTTrainer`], first install by 
 
 ```bash
 pip install liger-kernel
 ```
 
-2. Once installed, set `use_liger` in [`SFTConfig`]. No other changes are needed!
+2. Once installed, set `use_liger_kernel` in [`SFTConfig`]. No other changes are needed!
 
 ```python
 training_args = SFTConfig(
-  use_liger=True
+  use_liger_kernel=True
 )
 ```
 

tests/slow/test_sft_slow.py

@@ -412,7 +412,7 @@ def test_sft_trainer_with_liger(self, model_name, packing):
                 max_steps=2,
                 packing=packing,
                 max_length=self.max_length,
-                use_liger=True,
+                use_liger_kernel=True,
             )
 
             trainer = SFTTrainer(

trl/trainer/dpo_config.py

@@ -387,15 +387,15 @@ class DPOConfig(TrainingArguments):
     )
 
     # Deprecated parameters
-    use_num_logits_to_keep: bool = field(
-        default=False,
+    use_num_logits_to_keep: Optional[bool] = field(
+        default=None,
         metadata={"help": "Deprecated. Use `use_logits_to_keep` instead."},
     )
 
     def __post_init__(self):
         super().__post_init__()
 
-        if self.use_num_logits_to_keep:
+        if self.use_num_logits_to_keep is not None:
             warnings.warn(
                 "`use_num_logits_to_keep` is deprecated and will be remove in version 0.17.0. Use "
                 "`use_logits_to_keep` instead.",

trl/trainer/gkd_trainer.py

@@ -36,7 +36,7 @@
 )
 from transformers.trainer_callback import TrainerCallback
 from transformers.trainer_utils import EvalPrediction
-from transformers.utils import is_liger_kernel_available, is_peft_available
+from transformers.utils import is_peft_available
 
 from ..models import PreTrainedModelWrapper
 from ..models.utils import unwrap_model_for_generation
@@ -54,8 +54,6 @@
 if is_deepspeed_available():
     import deepspeed
 
-if is_liger_kernel_available():
-    from liger_kernel.transformers import AutoLigerKernelForCausalLM
 
 if is_peft_available():
     from peft import PeftConfig
@@ -119,10 +117,7 @@ def __init__(
             )
 
         if isinstance(teacher_model, str):
-            if args.use_liger:
-                teacher_model = AutoLigerKernelForCausalLM.from_pretrained(teacher_model, **teacher_model_init_kwargs)
-            else:
-                teacher_model = AutoModelForCausalLM.from_pretrained(teacher_model, **teacher_model_init_kwargs)
+            teacher_model = AutoModelForCausalLM.from_pretrained(teacher_model, **teacher_model_init_kwargs)
 
         # Disable dropout in the model
         if args.disable_dropout:

trl/trainer/sft_config.py

@@ -37,8 +37,6 @@ class SFTConfig(TrainingArguments):
         model_init_kwargs (`dict[str, Any]` or `None`, *optional*, defaults to `None`):
             Keyword arguments for [`~transformers.AutoModelForCausalLM.from_pretrained`], used when the `model`
             argument of the [`SFTTrainer`] is provided as a string.
-        use_liger (`bool`, *optional*, defaults to `False`):
-            Monkey patch the model with Liger kernels to increase throughput and reduce memory usage.
 
         > Parameters that control the data preprocessing
 
@@ -72,10 +70,6 @@ class SFTConfig(TrainingArguments):
             "the `SFTTrainer` is provided as a string."
         },
     )
-    use_liger: bool = field(
-        default=False,
-        metadata={"help": "Monkey patch the model with Liger kernels to increase throughput and reduce memory usage."},
-    )
 
     # Parameters that control the data preprocessing
     dataset_text_field: str = field(
@@ -123,25 +117,29 @@ class SFTConfig(TrainingArguments):
     )
 
     # Deprecated parameters
-    dataset_batch_size: int = field(
+    dataset_batch_size: Optional[int] = field(
         default=None,
         metadata={"help": "Deprecated. You can safely remove this parameter from your configuration."},
     )
-    num_of_sequences: int = field(
+    num_of_sequences: Optional[int] = field(
         default=None,
         metadata={
             "help": "Deprecated. Use `max_length` instead, which specifies the maximum length of the tokenized "
             "sequence, unlike `num_of_sequences`, which referred to string sequences."
         },
     )
-    chars_per_token: float = field(
+    chars_per_token: Optional[float] = field(
         default=None,
         metadata={"help": "Deprecated. If you want to customize the packing length, use `max_length`."},
     )
     max_seq_length: Optional[int] = field(
         default=None,
         metadata={"help": "Deprecated. Use `max_length` instead."},
     )
+    use_liger: Optional[bool] = field(
+        default=None,
+        metadata={"help": "Deprecated. Use `use_liger_kernel` instead."},
+    )
 
     def __post_init__(self):
         super().__post_init__()
@@ -174,3 +172,10 @@ def __post_init__(self):
                 DeprecationWarning,
             )
             self.max_length = self.max_seq_length
+
+        if self.use_liger is not None:
+            warnings.warn(
+                "`use_liger` is deprecated and will be remove in version 0.18.0. Use `use_liger_kernel` instead.",
+                DeprecationWarning,
+            )
+            self.use_liger_kernel = self.use_liger

trl/trainer/sft_trainer.py

@@ -40,7 +40,7 @@
 )
 from transformers.trainer_callback import TrainerCallback
 from transformers.trainer_utils import EvalPrediction
-from transformers.utils import is_liger_kernel_available, is_peft_available
+from transformers.utils import is_peft_available
 
 from ..data_utils import is_conversational, maybe_apply_chat_template, maybe_convert_to_chatml, pack_examples
 from .sft_config import SFTConfig
@@ -51,9 +51,6 @@
     import peft
     from peft import PeftConfig, PeftModel, get_peft_model, prepare_model_for_kbit_training
 
-if is_liger_kernel_available():
-    from liger_kernel.transformers import AutoLigerKernelForCausalLM
-
 if is_wandb_available():
     import wandb
 
@@ -264,12 +261,7 @@ def _create_model_from_path(self, model_path: str, args: SFTConfig) -> PreTraine
             model_init_kwargs["use_cache"] = False
 
         # Create model
-        if args.use_liger:
-            if not is_liger_kernel_available():
-                raise ImportError("Please install Liger-kernel for use_liger=True")
-            model = AutoLigerKernelForCausalLM.from_pretrained(model_path, **model_init_kwargs)
-        else:
-            model = AutoModelForCausalLM.from_pretrained(model_path, **model_init_kwargs)
+        model = AutoModelForCausalLM.from_pretrained(model_path, **model_init_kwargs)
         return model
 
     def _prepare_peft_model(self, model: PreTrainedModel, peft_config: Any, args: SFTConfig) -> PreTrainedModel:
@@ -457,7 +449,7 @@ def truncate(example, max_length):
                     **map_kwargs,
                 )
             # For Liger kernel, ensure only input_ids is present
-            if args.use_liger:
+            if args.use_liger_kernel:
                 dataset = dataset.select_columns("input_ids")
 
         return dataset
@@ -471,7 +463,7 @@ def compute_loss(self, model, inputs, return_outputs=False, num_items_in_batch=N
         )
 
         # Compute token accuracy if we have labels and if the model is not using Liger (no logits)
-        if "labels" in inputs and not self.args.use_liger:
+        if "labels" in inputs and not self.args.use_liger_kernel:
             shift_logits = outputs.logits[..., :-1, :].contiguous()
             shift_labels = inputs["labels"][..., 1:].contiguous()