diff --git a/ext/flan_t5.py b/ext/flan_t5.py
index 8025c34..353e41a 100644
--- a/ext/flan_t5.py
+++ b/ext/flan_t5.py
@@ -7,7 +7,7 @@
 class FlanT5(BaseLM):
 
     def __init__(self, model_name="google/flan-t5-base", temp=0.1, device='cuda', max_length=None, use_bf16=False, **kwargs):
-        super(FlanT5, self).__init__(name=model_name)
+        super(FlanT5, self).__init__(name=model_name, **kwargs)
         self.__device = device
         self.__max_length = 512 if max_length is None else max_length
         self.__model = T5ForConditionalGeneration.from_pretrained(
diff --git a/ext/gemma.py b/ext/gemma.py
index 5a32ddc..81d8418 100644
--- a/ext/gemma.py
+++ b/ext/gemma.py
@@ -7,8 +7,8 @@
 class Gemma(BaseLM):
 
     def __init__(self, model_name="google/gemma-7b-it", temp=0.1, device='cuda',
-                 max_length=None, api_token=None, use_bf16=False):
-        super(Gemma, self).__init__(name=model_name)
+                 max_length=None, api_token=None, use_bf16=False, **kwargs):
+        super(Gemma, self).__init__(name=model_name, **kwargs)
         self.__device = device
         self.__max_length = 1024 if max_length is None else max_length
         self.__model = AutoModelForCausalLM.from_pretrained(
diff --git a/ext/llama32.py b/ext/llama32.py
index d9b099d..b94cf50 100644
--- a/ext/llama32.py
+++ b/ext/llama32.py
@@ -7,8 +7,8 @@
 class Llama32(BaseLM):
 
     def __init__(self, model_name="meta-llama/Llama-3.2-3B-Instruct", api_token=None,
-                 temp=0.1, device='cuda', max_length=256, use_bf16=False):
-        super(Llama32, self).__init__(name=model_name)
+                 temp=0.1, device='cuda', max_length=256, use_bf16=False, **kwargs):
+        super(Llama32, self).__init__(name=model_name, **kwargs)
 
         if use_bf16:
             print("Warning: Experimental mode with bf-16!")
diff --git a/ext/microsoft_phi_2.py b/ext/microsoft_phi_2.py
index 3d5aae1..e66b849 100644
--- a/ext/microsoft_phi_2.py
+++ b/ext/microsoft_phi_2.py
@@ -11,8 +11,8 @@ class MicrosoftPhi2(BaseLM):
     """ https://huggingface.co/microsoft/phi-2
     """
 
-    def __init__(self, model_name="microsoft/phi-2", device='cuda', max_length=None, use_bf16=False):
-        super(MicrosoftPhi2, self).__init__(model_name)
+    def __init__(self, model_name="microsoft/phi-2", device='cuda', max_length=None, use_bf16=False, **kwargs):
+        super(MicrosoftPhi2, self).__init__(model_name, **kwargs)
 
         # Default parameters.
         kwargs = {"device_map": device, "trust_remote_code": True}
diff --git a/ext/mistral.py b/ext/mistral.py
index 20fe828..04d8ad9 100644
--- a/ext/mistral.py
+++ b/ext/mistral.py
@@ -7,9 +7,9 @@
 class Mistral(BaseLM):
 
     def __init__(self, model_name="mistralai/Mistral-7B-Instruct-v0.1", temp=0.1, device='cuda', max_length=None,
-                 use_bf16=False):
+                 use_bf16=False, **kwargs):
         assert(isinstance(max_length, int) or max_length is None)
-        super(Mistral, self).__init__(name=model_name)
+        super(Mistral, self).__init__(name=model_name, **kwargs)
 
         if use_bf16:
             print("Warning: Experimental mode with bf-16!")
diff --git a/ext/openai_gpt.py b/ext/openai_gpt.py
index f95d7a1..67ff413 100644
--- a/ext/openai_gpt.py
+++ b/ext/openai_gpt.py
@@ -5,9 +5,9 @@
 class OpenAIGPT(BaseLM):
 
     def __init__(self, api_key, model_name="gpt-4-1106-preview", temp=0.1, max_tokens=None, assistant_prompt=None,
-                 freq_penalty=0.0, kwargs=None):
+                 freq_penalty=0.0, **kwargs):
         assert(isinstance(assistant_prompt, str) or assistant_prompt is None)
-        super(OpenAIGPT, self).__init__(name=model_name)
+        super(OpenAIGPT, self).__init__(name=model_name, **kwargs)
 
         # dynamic import of the OpenAI library.
         OpenAI = auto_import("openai._client.OpenAI", is_class=False)