FEAT / Trainer: Add adamw 4bit optimizer (#31865)

* add 4bit optimizer

* style

* fix msg

* style

* add qgalore

* Revert "add qgalore"

This reverts commit 25278e8.

* style

* version check

Author: SunMarc

src/transformers/trainer.py

@@ -168,6 +168,7 @@
     is_torch_npu_available,
     is_torch_xla_available,
     is_torch_xpu_available,
+    is_torchao_available,
     logging,
     strtobool,
 )
@@ -1451,7 +1452,23 @@ def optimizer_hook(param):
                     "gradient_clipping": float(optim_args.get("gradient_clipping", 1.0)),
                 }
             )
+        elif args.optim == OptimizerNames.ADAMW_TORCH_4BIT:
+            if not is_torchao_available() or version.parse(importlib.metadata.version("torchao")) < version.parse(
+                "0.4.0"
+            ):
+                raise ImportError(
+                    "You need to have `torchao>=0.4.0` in order to use torch 4-bit optimizers."
+                    "Install it with `pip install torchao` or follow the instructions here: https://github.com/pytorch/ao"
+                )
+            if version.parse(importlib.metadata.version("torch")) < version.parse("2.3"):
+                raise ImportError(
+                    "You need to have `torch>=2.3` in order to use torch 4-bit optimizers. "
+                    "Install it with `pip install --upgrade torch`"
+                )
+            from torchao.prototype.low_bit_optim import AdamW4bit
 
+            optimizer_cls = AdamW4bit
+            optimizer_kwargs.update(adam_kwargs)
         else:
             raise ValueError(f"Trainer cannot instantiate unsupported optimizer: {args.optim}")
         return optimizer_cls, optimizer_kwargs

src/transformers/training_args.py

@@ -154,6 +154,7 @@ class OptimizerNames(ExplicitEnum):
     ADAMW_APEX_FUSED = "adamw_apex_fused"
     ADAFACTOR = "adafactor"
     ADAMW_ANYPRECISION = "adamw_anyprecision"
+    ADAMW_TORCH_4BIT = "adamw_torch_4bit"
     SGD = "sgd"
     ADAGRAD = "adagrad"
     ADAMW_BNB = "adamw_bnb_8bit"

tests/trainer/test_trainer.py

@@ -99,6 +99,7 @@
     is_apex_available,
     is_bitsandbytes_available,
     is_safetensors_available,
+    is_torchao_available,
     is_torchdistx_available,
 )
 from transformers.utils.hp_naming import TrialShortNamer
@@ -4210,6 +4211,16 @@ def hp_name(trial):
                 dict(default_adam_kwargs, **default_anyprecision_kwargs),
             )
         )
+    if is_torchao_available():
+        import torchao
+
+        optim_test_params.append(
+            (
+                TrainingArguments(optim=OptimizerNames.ADAMW_TORCH_4BIT, output_dir="None"),
+                torchao.prototype.low_bit_optim.AdamW4bit,
+                default_adam_kwargs,
+            )
+        )
 
 
 @require_torch