[BUG] PowerPaintV1 Cannot run in batch processing

[HelloRheina]

Model
PowerPaintv1

Describe the bug
│ │
│ 56 │ │ │ self.model.enable_sequential_cpu_offload(gpu_id=0) │
│ 57 │ │ else: │
│ 58 │ │ │ self.model = self.model.to(device) │
│ ❱ 59 │ │ │ if kwargs["sd_cpu_textencoder"]: │
│ 60 │ │ │ │ logger.info("Run Stable Diffusion TextEncoder on CPU") │
│ 61 │ │ │ │ self.model.text_encoder = CPUTextEncoderWrapper( │
│ 62 │ │ │ │ │ self.model.text_encoder, torch_dtype │
╰──────────────────────────────────────────────────────────────────────────────╯
KeyError: 'sd_cpu_textencoder'

I am running PoerPaintV1 using batch processing
$ iopaint run --device=cuda --model=Sanster/PowerPaint-V1-stable-diffusion-inpainting --image=./image --mask=./mask --output=./powerpaint

It runs when I use the localhost web UI but not with batch processing in terminal

This is my env

• Platform: Linux-6.8.0-49-generic-x86_64-with-glibc2.39
• Python version: 3.9.20
• torch: 2.5.1
• torchvision: 0.20.1
• Pillow: 9.5.0
• diffusers: 0.27.0
• transformers: 4.28.0
• opencv-python: 4.10.0.84
• accelerate: 1.1.1
• iopaint: 1.5.3
• rembg: N/A

Screenshots
If applicable, add screenshots to help explain your problem.
Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s]Some weights of StableDiffusionSafetyChecker were not initialized from the model checkpoint at /home/rei/.cache/huggingface/hub/models--Sanster--PowerPaint-V1-stable-diffusion-inpainting/snapshots/4104236baefdb2a1dd0f32812d8edbbfc4efd164/safety_checker and are newly initialized: ['vision_model.vision_model.embeddings.position_ids']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
Loading pipeline components...: 100%|█████████████| 7/7 [00:00<00:00, 8.04it/s]
╭───────────────────── Traceback (most recent call last) ──────────────────────╮
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ cli.py:94 in run │
│ │
│ 91 │ │
│ 92 │ from iopaint.batch_processing import batch_inpaint │
│ 93 │ │
│ ❱ 94 │ batch_inpaint(model, device, image, mask, output, config, concat) │
│ 95 │
│ 96 │
│ 97 @typer_app.command(help="Start IOPaint server") │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ batch_processing.py:71 in batch_inpaint │
│ │
│ 68 │ │ │ inpaint_request = InpaintRequest(**json.load(f)) │
│ 69 │ │ logger.info(f"Using config: {inpaint_request}") │
│ 70 │ │
│ ❱ 71 │ model_manager = ModelManager(name=model, device=device) │
│ 72 │ first_mask = list(mask_paths.values())[0] │
│ 73 │ │
│ 74 │ console = Console() │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ model_manager.py:39 in init │
│ │
│ 36 │ │ │
│ 37 │ │ self.enable_powerpaint_v2 = kwargs.get("enable_powerpaint_v2", │
│ 38 │ │ │
│ ❱ 39 │ │ self.model = self.init_model(name, device, **kwargs) │
│ 40 │ │
│ 41 │ @Property │
│ 42 │ def current_model(self) -> ModelInfo: │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ model_manager.py:72 in init_model │
│ │
│ 69 │ │ │ return PowerPaintV2(device, **kwargs) │
│ 70 │ │ │
│ 71 │ │ if model_info.name in models: │
│ ❱ 72 │ │ │ return models[name](device, **kwargs) │
│ 73 │ │ │
│ 74 │ │ if model_info.model_type in [ │
│ 75 │ │ │ ModelType.DIFFUSERS_SD_INPAINT, │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ model/base.py:277 in init │
│ │
│ 274 │ def init(self, device, **kwargs): │
│ 275 │ │ self.model_info = kwargs["model_info"] │
│ 276 │ │ self.model_id_or_path = self.model_info.path │
│ ❱ 277 │ │ super().init(device, **kwargs) │
│ 278 │ │
│ 279 │ @torch.no_grad() │
│ 280 │ def call(self, image, mask, config: InpaintRequest): │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ model/base.py:35 in init │
│ │
│ 32 │ │ """ │
│ 33 │ │ device = switch_mps_device(self.name, device) │
│ 34 │ │ self.device = device │
│ ❱ 35 │ │ self.init_model(device, **kwargs) │
│ 36 │ │
│ 37 │ @abc.abstractmethod │
│ 38 │ def init_model(self, device, **kwargs): ... │
│ │
│ /home/rei/anaconda3/envs/gen_left_right/lib/python3.9/site-packages/iopaint/ │
│ model/power_paint/power_paint.py:59 in init_model │
│ │
│ 56 │ │ │ self.model.enable_sequential_cpu_offload(gpu_id=0) │
│ 57 │ │ else: │
│ 58 │ │ │ self.model = self.model.to(device) │
│ ❱ 59 │ │ │ if kwargs["sd_cpu_textencoder"]: │
│ 60 │ │ │ │ logger.info("Run Stable Diffusion TextEncoder on CPU") │
│ 61 │ │ │ │ self.model.text_encoder = CPUTextEncoderWrapper( │
│ 62 │ │ │ │ │ self.model.text_encoder, torch_dtype │
╰──────────────────────────────────────────────────────────────────────────────╯

System Info
This is my env

• Platform: Linux-6.8.0-49-generic-x86_64-with-glibc2.39
• Python version: 3.9.20
• torch: 2.5.1
• torchvision: 0.20.1
• Pillow: 9.5.0
• diffusers: 0.27.0
• transformers: 4.28.0
• opencv-python: 4.10.0.84
• accelerate: 1.1.1
• iopaint: 1.5.3
• rembg: N/A