dx8152

/

Flux2-Klein-9B-Consistency

like 460

I used it on diffuers
transformer = Flux2Transformer2DModel.from_pretrained(
self.model_dir, subfolder="transformer", torch_dtype=torch.bfloat16,
device_map={"": 0},
)
pipe = replica_cls.from_pretrained(
self.model_dir, transformer=transformer, text_encoder=None,
torch_dtype=torch.bfloat16,
)
pipe.vae.to("cuda:0")
pipe.load_lora_weights(self.model_dir, weight_name="Flux2-Klein-9B-consistency-V2.safetensors")

The person generated and the one I uploaded are completely different. Can you please explain why this is happening