use default torch fused adamw optimizer as default as adamw_hf is deprecated (#2425)

* use default torch fused adamw optimizer as default as adamw_hf is deprecated

* make sure to have latest packaging installed

* bump packagingin requirements.txt too

pyproject.toml

@@ -1,5 +1,5 @@
 [build-system]
-requires = ["setuptools>=64", "wheel", "setuptools_scm>=8"]
+requires = ["setuptools>=64", "wheel", "setuptools_scm>=8", "packaging>=24.2"]
 build-backend = "setuptools.build_meta"
 
 [project]

requirements.txt

@@ -10,7 +10,7 @@ autoawq==0.2.7.post3
 liger-kernel==0.5.3
 # END section
 
-packaging==23.2
+packaging==24.2
 
 peft==0.14.0
 transformers==4.49.0

src/axolotl/utils/config/models/input/v0_4_1/__init__.py

@@ -507,7 +507,7 @@ class HyperparametersConfig(BaseModel):
     weight_decay: Optional[float] = 0.0
     optimizer: Optional[
         Union[OptimizerNames, CustomSupportedOptimizers]
-    ] = OptimizerNames.ADAMW_HF
+    ] = OptimizerNames.ADAMW_TORCH_FUSED
     optim_args: Optional[Union[str, Dict[str, Any]]] = Field(
         default=None,
         json_schema_extra={"description": "Optional arguments to supply to optimizer."},