diff --git a/litgpt/finetune/adapter.py b/litgpt/finetune/adapter.py
index aa02977fb9..2ec71784e7 100644
--- a/litgpt/finetune/adapter.py
+++ b/litgpt/finetune/adapter.py
@@ -145,7 +145,6 @@ def main(
 
     model = fabric.setup_module(model)
 
-    trainable_params = [p for p in model.parameters() if p.requires_grad]
     if isinstance(fabric.strategy.precision, BitsandbytesPrecision):
         import bitsandbytes as bnb
 
@@ -153,7 +152,7 @@ def main(
     else:
         optimizer_cls = torch.optim.AdamW
     optimizer = optimizer_cls(
-        trainable_params, lr=train.learning_rate, weight_decay=train.weight_decay, betas=(train.beta1, train.beta2)
+        model.parameters(), lr=train.learning_rate, weight_decay=train.weight_decay, betas=(train.beta1, train.beta2)
     )
     optimizer = fabric.setup_optimizers(optimizer)
     scheduler = get_lr_scheduler(optimizer, warmup_steps=train.lr_warmup_steps, max_steps=lr_max_steps)
diff --git a/litgpt/finetune/adapter_v2.py b/litgpt/finetune/adapter_v2.py
index 197bcd0bba..86526a58e5 100644
--- a/litgpt/finetune/adapter_v2.py
+++ b/litgpt/finetune/adapter_v2.py
@@ -145,7 +145,6 @@ def main(
 
     model = fabric.setup_module(model)
 
-    trainable_params = [p for p in model.parameters() if p.requires_grad]
     if isinstance(fabric.strategy.precision, BitsandbytesPrecision):
         import bitsandbytes as bnb
 
@@ -153,7 +152,7 @@ def main(
     else:
         optimizer_cls = torch.optim.AdamW
     optimizer = optimizer_cls(
-        trainable_params, lr=train.learning_rate, weight_decay=train.weight_decay, betas=(train.beta1, train.beta2)
+        model.parameters(), lr=train.learning_rate, weight_decay=train.weight_decay, betas=(train.beta1, train.beta2)
     )
     optimizer = fabric.setup_optimizers(optimizer)
     scheduler = get_lr_scheduler(optimizer, warmup_steps=train.lr_warmup_steps, max_steps=lr_max_steps)