disable quantization

app.py

@@ -39,7 +39,7 @@ print(">>> MODEL CACHE PATH:", MODEL_CACHE, os.listdir(MODEL_CACHE))
 
 device = "gpu" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
 dtype = torch.float16 if device in ("mps", "gpu") else torch.float32
-bnb_config = BitsAndBytesConfig(load_in_8bit=True)
+# bnb_config = BitsAndBytesConfig(load_in_8bit=True)
 
 
 # If lower memory usage needed: 
@@ -77,7 +77,7 @@ def load_model():
         local_files_only=True,
         torch_dtype=dtype,
         trust_remote_code=True,
-        quantization_config=bnb_config,
+        # quantization_config=bnb_config,
         no_split_module_classes=["Block"],
         device_map="auto"
     ).to(device).eval()