diff --git a/modeling/inference_models/generic_hf_torch/class.py b/modeling/inference_models/generic_hf_torch/class.py
index b51d8f66..9a59650e 100644
--- a/modeling/inference_models/generic_hf_torch/class.py
+++ b/modeling/inference_models/generic_hf_torch/class.py
@@ -80,7 +80,6 @@ class model_backend(HFTorchInferenceModel):
         if self.use_4_bit:
             self.lazy_load = False
             tf_kwargs.update({
-                "load_in_4bit": True,
                 "quantization_config":BitsAndBytesConfig(
                     load_in_4bit=True,
                     bnb_4bit_compute_dtype=torch.float16,