Fix idefics.

server/text_generation_server/models/idefics.py

@@ -33,6 +33,7 @@ class IDEFICSSharded(IdeficsCausalLM):
         dtype: Optional[torch.dtype] = None,
         trust_remote_code: bool = False,
     ):
+        self.quantize = quantize
         self.process_group, rank, world_size = initialize_torch_distributed()
         if torch.cuda.is_available():
             device = torch.device(f"cuda:{rank}")

server/text_generation_server/models/idefics_causal_lm.py

@@ -580,6 +580,7 @@ class IdeficsCausalLM(Model):
         dtype: Optional[torch.dtype] = None,
         trust_remote_code: bool = False,
     ):
+        self.quantize = quantize
         from text_generation_server.models.custom_modeling.idefics_modeling import (
             IdeficsForVisionText2Text,
         )