diff --git a/src/axolotl/monkeypatch/models/llama/modeling_llama.py b/src/axolotl/monkeypatch/models/llama/modeling_llama.py
index c0a20fbd3..94a4f6580 100644
--- a/src/axolotl/monkeypatch/models/llama/modeling_llama.py
+++ b/src/axolotl/monkeypatch/models/llama/modeling_llama.py
@@ -107,19 +107,19 @@ class AxolotlLlamaForCausalLM(LlamaForCausalLM):
             compile(forward_source, "<forward>", "exec"), self
         )
 
-    def enable_act_offloading(self):
-        self.act_offloading_ctx_manager = OffloadActivations()
-
-        forward_source = inspect.getsource(self.forward)
+    @classmethod
+    def enable_act_offloading(cls):
+        forward_source = inspect.getsource(cls.forward)
         forward_source = forward_source.replace(
             HF_MODEL_OUTPUTS, PATCHED_HF_MODEL_OUTPUTS
         )
         # replace forward method with patched version
-        self.forward = types.MethodType(
-            compile(forward_source, "<llama_forward_w_act_offloading>", "exec"), self
+        cls.forward = types.MethodType(
+            compile(forward_source, "<llama_forward_w_act_offloading>", "exec"), cls
         )
 
-    def enable_liger_fce(self, enable_act_offloading=True):
+    @classmethod
+    def enable_liger_fce(cls, enable_act_offloading=True):
         from liger_kernel.transformers.model.llama import (
             lce_forward as llama_lce_forward,
         )
@@ -128,16 +128,17 @@ class AxolotlLlamaForCausalLM(LlamaForCausalLM):
             lce_source = inspect.getsource(llama_lce_forward)
             lce_source = lce_source.replace(LCE_MODEL_OUTPUTS, PATCHED_LCE_OUTPUTS)
             # replace forward method with patched version
-            self.forward = types.MethodType(
+            cls.forward = types.MethodType(
                 compile(lce_source, "<llama_lce_forward_w_act_offloading>", "exec"),
-                self,
+                cls,
             )
         else:
-            self.forward = types.methodType(llama_lce_forward, self)
+            cls.forward = types.methodType(llama_lce_forward, cls)
 
-    def patch_hf_ga(self):
+    @classmethod
+    def patch_hf_ga(cls):
         # bugfix patch for gradient accumulation
-        forward_source = inspect.getsource(self.forward)
+        forward_source = inspect.getsource(cls.forward)
         forward_source = forward_source.replace(
             HF_GA_FORWARD_1, PATCHED_HF_GA_FORWARD_1
         )
@@ -145,8 +146,8 @@ class AxolotlLlamaForCausalLM(LlamaForCausalLM):
             HF_GA_FORWARD_2, PATCHED_HF_GA_FORWARD_2
         )
         # replace forward method with patched version
-        self.forward = types.MethodType(
-            compile(forward_source, "<llama_forward_ga_fix>", "exec"), self
+        cls.forward = types.MethodType(
+            compile(forward_source, "<llama_forward_ga_fix>", "exec"), cls
         )
 
 
@@ -155,3 +156,5 @@ def replace_auto_model():
     from transformers.models.auto import MODEL_FOR_CAUSAL_LM_MAPPING
 
     MODEL_FOR_CAUSAL_LM_MAPPING[LlamaConfig] = AxolotlLlamaForCausalLM
+
+    return AxolotlLlamaForCausalLM
diff --git a/src/axolotl/utils/models.py b/src/axolotl/utils/models.py
index 4db9107c7..b44e1ec38 100644
--- a/src/axolotl/utils/models.py
+++ b/src/axolotl/utils/models.py
@@ -380,6 +380,15 @@ class ModelLoader:
         plugin_manager = PluginManager.get_instance()
         plugin_manager.pre_model_load(self.cfg)
 
+        if self.cfg.model_config_type == "llama":
+            from axolotl.monkeypatch.models.llama.modeling_llama import replace_auto_model
+
+            AxolotlLlamaForCausalLM = replace_auto_model()
+
+            AxolotlLlamaForCausalLM.patch_hf_ga()
+            if self.cfg.activation_offloading:
+                AxolotlLlamaForCausalLM.enable_act_offloading()
+
         if self.cfg.fsdp:
             from axolotl.monkeypatch.trainer_fsdp_optim import (
                 patch_training_loop_for_fsdp,