Fix adopt descriptions

Author: rwightman

timm/optim/_optim_factory.py

@@ -432,17 +432,17 @@ def _register_adam_variants(registry: OptimizerRegistry) -> None:
         OptimInfo(
             name='adafactorbv',
             opt_class=AdafactorBigVision,
-            description='Big Vision variant of Adafactor with factored gradients, half precision momentum.',
+            description='Big Vision variant of Adafactor with factored gradients, half precision momentum',
         ),
         OptimInfo(
             name='adopt',
             opt_class=Adopt,
-            description='Memory-efficient implementation of Adam with factored gradients',
+            description='Modified Adam that can converge with any β2 with the optimal rate',
         ),
         OptimInfo(
             name='adoptw',
             opt_class=Adopt,
-            description='Memory-efficient implementation of Adam with factored gradients',
+            description='Modified AdamW (decoupled decay) that can converge with any β2 with the optimal rate',
             defaults={'decoupled': True}
         ),
     ]

timm/optim/adopt.py

@@ -51,6 +51,10 @@ def _get_value(x):
 
 
 class Adopt(Optimizer):
+    """
+    ADOPT: Modified Adam Can Converge with Any β2 with the Optimal Rate: https://arxiv.org/abs/2411.02853
+
+    """
     def __init__(
             self,
             params,