Fix for --nobreakmodel forcing CPU

Put importing of colab packages into a if function so it doesn't error out

modeling/inference_models/generic_hf_torch/class.py

@@ -250,7 +250,7 @@ class model_backend(HFTorchInferenceModel):
 
         
         if utils.koboldai_vars.hascuda:
-            if self.usegpu:
+            if self.usegpu or self.nobreakmodel:
                 # Use just VRAM
                 self.model = self.model.half().to(utils.koboldai_vars.gpu_device)
             elif self.breakmodel: