diff --git a/unsloth/models/gemma.py b/unsloth/models/gemma.py
index 7173c03495..55a8c8697f 100644
--- a/unsloth/models/gemma.py
+++ b/unsloth/models/gemma.py
@@ -442,10 +442,10 @@ class FastGemmaModel(FastLlamaModel):
         return
 
     @staticmethod
-    def post_patch(model, tokenizer):
+    def post_patch(model, tokenizer, correct_dtype = None):
         # Gemma does not downcast RoPE
         model, tokenizer = patch_model_and_tokenizer(
-            model, tokenizer, downcast_rope = False
+            model, tokenizer, downcast_rope = False, correct_dtype = correct_dtype
         )
 
         # Add 1 to weight
diff --git a/unsloth/models/gemma2.py b/unsloth/models/gemma2.py
index 16d04955d3..03e77f6504 100644
--- a/unsloth/models/gemma2.py
+++ b/unsloth/models/gemma2.py
@@ -613,10 +613,10 @@ class FastGemma2Model(FastLlamaModel):
         return
 
     @staticmethod
-    def post_patch(model, tokenizer):
+    def post_patch(model, tokenizer, correct_dtype = None):
         # Gemma does not downcast RoPE
         model, tokenizer = patch_model_and_tokenizer(
-            model, tokenizer, downcast_rope = False
+            model, tokenizer, downcast_rope = False, correct_dtype = correct_dtype
         )
 
         # Add 1 to weight
diff --git a/unsloth/models/granite.py b/unsloth/models/granite.py
index aae746aed1..168df90f4c 100644
--- a/unsloth/models/granite.py
+++ b/unsloth/models/granite.py
@@ -542,7 +542,7 @@ class FastGraniteModel(FastLlamaModel):
         return
 
     @staticmethod
-    def post_patch(model, tokenizer):
+    def post_patch(model, tokenizer, correct_dtype = None):
         # Torch.compile fails on embedding matrix??
         # Workaround randomnly fixes it for torch versions < 2.2
         model.model.embed_tokens = torch.nn.Embedding.from_pretrained(
diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 61771e4567..c1e9110759 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -2483,7 +2483,9 @@ class FastLlamaModel:
         )
 
         model, tokenizer = patch_tokenizer(model, tokenizer)
-        model, tokenizer = model_patcher.post_patch(model, tokenizer)
+        model, tokenizer = model_patcher.post_patch(
+            model, tokenizer, correct_dtype = dtype
+        )
 
         # Patch up QKV / O and MLP
         for idx, layer in enumerate(model.model.layers):
@@ -2666,9 +2668,9 @@ class FastLlamaModel:
         return model, tokenizer
 
     @staticmethod
-    def post_patch(model, tokenizer):
+    def post_patch(model, tokenizer, correct_dtype = None):
         model, tokenizer = patch_model_and_tokenizer(
-            model, tokenizer, downcast_rope = True
+            model, tokenizer, downcast_rope = True, correct_dtype = correct_dtype
         )
         return model, tokenizer
 
diff --git a/unsloth/models/rl.py b/unsloth/models/rl.py
index 6cb12f6a12..67721d7531 100755
--- a/unsloth/models/rl.py
+++ b/unsloth/models/rl.py
@@ -1309,6 +1309,18 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
                 flags = re.DOTALL,
             )
 
+    # Remove TRL's unconditional bfloat16 cast of trainable params (added in
+    # TRL 0.26.0). TRL hardcodes bfloat16 for QLoRA per the original paper's
+    # recommendation, but this is wrong: it ignores the user's requested dtype
+    # and breaks GradScaler when training with fp16=True. Unsloth already
+    # handles adapter dtype correctly via patch_model_and_tokenizer, so the
+    # entire block is unnecessary. For GRPOTrainer the enclosing peft init
+    # block is already removed above, making this a no-op for GRPO.
+    RLTrainer_source = RLTrainer_source.replace(
+        'if getattr(model, "is_loaded_in_4bit", False) or getattr(model, "is_loaded_in_8bit", False):',
+        "if False:",
+    )
+
     if RLTrainer_name == "SFTTrainer":
         original_text = 'self._signature_columns = ["input_ids", "attention_mask", "completion_mask"]'
         new_text = 'self._signature_columns = ["input_ids", "attention_mask", "completion_mask","labels"]'