Set Flux models to eval mode to restore FP8 performance

Fixes performance regression introduced in Comfy-Org#9854 where the fp8_linear
optimization path was being skipped during inference because Flux models
were not explicitly set to eval mode.

Author: mturnshek

comfy/model_base.py

@@ -834,6 +834,7 @@ def extra_conds(self, **kwargs):
 class Flux(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLUX, device=None, unet_model=comfy.ldm.flux.model.Flux):
         super().__init__(model_config, model_type, device=device, unet_model=unet_model)
+        self.diffusion_model.eval().requires_grad_(False)
         self.memory_usage_factor_conds = ("ref_latents",)
 
     def concat_cond(self, **kwargs):