can't run on macos m1 max, os version: 26.3

[qzd1989]

I can't run it, the output below:

kiwi@Delphi ~/Documents/Code % git clone https://github.com/D-Ogi/WatermarkRemover-AI.git
Cloning into 'WatermarkRemover-AI'...
remote: Enumerating objects: 185, done.
remote: Counting objects: 100% (60/60), done.
remote: Compressing objects: 100% (31/31), done.
remote: Total 185 (delta 45), reused 29 (delta 29), pack-reused 125 (from 2)
Receiving objects: 100% (185/185), 9.18 MiB | 314.00 KiB/s, done.
Resolving deltas: 100% (71/71), done.
kiwi@Delphi ~/Documents/Code % cd WatermarkRemover-AI
kiwi@Delphi ~/Documents/Code/WatermarkRemover-AI % chmod +x setup.sh
kiwi@Delphi ~/Documents/Code/WatermarkRemover-AI % ./setup.sh

  =============================================
     WatermarkRemover-AI Setup (Linux/macOS)
  =============================================

  [?] Are you in China? (y/n)
      This will use faster mirrors for downloads
      y
  [OK] Using China mirrors (Tsinghua PyPI + HF-Mirror)

  [*] Detected macOS
  [OK] Found python3.12 (version 3.12)
  [*] Creating virtual environment...
  [OK] Virtual environment created
  [*] Upgrading pip...
  [*] Installing PyTorch...
  [OK] PyTorch installed (MPS support on Apple Silicon)
  [*] Installing other dependencies...
  [*] Installing iopaint...
  [*] Installing iopaint dependencies...
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
iopaint 1.6.0 requires accelerate, which is not installed.
iopaint 1.6.0 requires controlnet-aux==0.0.3, which is not installed.
iopaint 1.6.0 requires fastapi==0.108.0, which is not installed.
iopaint 1.6.0 requires gradio==4.21.0, which is not installed.
iopaint 1.6.0 requires opencv-python, which is not installed.
iopaint 1.6.0 requires peft==0.7.1, which is not installed.
iopaint 1.6.0 requires piexif==1.1.3, which is not installed.
iopaint 1.6.0 requires python-multipart, which is not installed.
iopaint 1.6.0 requires python-socketio==5.7.2, which is not installed.
iopaint 1.6.0 requires typer-config==1.4.0, which is not installed.
iopaint 1.6.0 requires uvicorn, which is not installed.
iopaint 1.6.0 requires diffusers==0.27.2, but you have diffusers 0.37.0 which is incompatible.
iopaint 1.6.0 requires huggingface-hub==0.25.2, but you have huggingface-hub 1.7.2 which is incompatible.
iopaint 1.6.0 requires Pillow==9.5.0, but you have pillow 12.1.1 which is incompatible.
  [OK] Dependencies installed
  [*] Downloading LaMA model (~196MB)...
  [OK] LaMA model already exists
  [*] Downloading Florence-2 model (~1.5GB)...
      Using HF-Mirror for faster download in China
/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/huggingface_hub/utils/_validators.py:206: UserWarning: The `local_dir_use_symlinks` argument is deprecated and ignored in `snapshot_download`. Downloading to a local directory does not use symlinks anymore.
  warnings.warn(
Fetching 13 files: 100%|█████████████████████████████████████████████████████████████| 13/13 [00:00<00:00, 213827.26it/s]
Download complete: : 0.00B [00:00, ?B/s]                                                          | 0/13 [00:00<?, ?it/s]

  =============================================
     Setup complete!
  =============================================

  To run the app:
    source venv/bin/activate
    python remwmgui.py

  Or for CLI:
    source venv/bin/activate
    python remwm.py input.png output/

  Launch now? (y/n) n
kiwi@Delphi ~/Documents/Code/WatermarkRemover-AI % source venv/bin/activate
((venv) ) kiwi@Delphi ~/Documents/Code/WatermarkRemover-AI % python remwm.py ~/Desktop/input.jpeg ~/Desktop/output/
/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/iopaint/model/ldm.py:279: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
  @torch.cuda.amp.autocast()
Using device: cpu
Loading weights: 100%|███████████████████████████████████████████████████████████████| 917/917 [00:00<00:00, 6005.63it/s]
Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.
The image processor of type `CLIPImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`.
2026-03-21 22:20:27.528 | INFO     | __main__:main:663 - Florence-2 Model loaded
2026-03-21 22:20:27.533 | INFO     | iopaint.model_manager:init_model:47 - Loading model: lama
2026-03-21 22:20:27.533 | INFO     | iopaint.helper:load_jit_model:107 - Loading model from: /Users/kiwi/.cache/torch/hub/checkpoints/big-lama.pt
2026-03-21 22:20:27.830 | INFO     | __main__:main:667 - LaMa model loaded
The following generation flags are not valid and may be ignored: ['early_stopping']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
Traceback (most recent call last):
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/remwm.py", line 705, in <module>
    main()
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/click/core.py", line 1485, in __call__
    return self.main(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/click/core.py", line 1406, in main
    rv = self.invoke(ctx)
         ^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/click/core.py", line 1269, in invoke
    return ctx.invoke(self.callback, **ctx.params)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/click/core.py", line 824, in invoke
    return callback(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/remwm.py", line 701, in main
    handle_one(input_path, output_file, florence_model, florence_processor, model_manager, device, transparent, max_bbox_percent, force_format, overwrite, detection_prompt, detection_skip, fade_in, fade_out)
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/remwm.py", line 525, in handle_one
    mask_image = get_watermark_mask(image, florence_model, florence_processor, device, max_bbox_percent, detection_prompt)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/remwm.py", line 107, in get_watermark_mask
    parsed_answer = identify(task_prompt, image, detection_prompt, model, processor, device)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/remwm.py", line 82, in identify
    generated_ids = model.generate(
                    ^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/generation/utils.py", line 2438, in generate
    model_kwargs = self._prepare_encoder_decoder_kwargs_for_generation(
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/models/florence2/modeling_florence2.py", line 1025, in _prepare_encoder_decoder_kwargs_for_generation
    image_features = self.get_image_features(pixel_values, return_dict=True).pooler_output
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/models/florence2/modeling_florence2.py", line 859, in get_image_features
    return self.model.get_image_features(pixel_values=pixel_values, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/utils/generic.py", line 843, in wrapper
    output = func(self, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/models/florence2/modeling_florence2.py", line 697, in get_image_features
    image_outputs = self.vision_tower(pixel_values, return_dict=True, **kwargs)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1776, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1787, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/utils/generic.py", line 917, in wrapper
    output = func(self, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/utils/output_capturing.py", line 253, in wrapper
    outputs = func(self, *args, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/models/florence2/modeling_florence2.py", line 562, in forward
    hidden_states = conv(hidden_states)
                    ^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1776, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1787, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/transformers/models/florence2/modeling_florence2.py", line 191, in forward
    hidden_states = self.conv(hidden_states)
                    ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1776, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1787, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/conv.py", line 553, in forward
    return self._conv_forward(input, self.weight, self.bias)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kiwi/Documents/Code/WatermarkRemover-AI/venv/lib/python3.12/site-packages/torch/nn/modules/conv.py", line 548, in _conv_forward
    return F.conv2d(
           ^^^^^^^^^
RuntimeError: Input type (float) and bias type (c10::Half) should be the same

[FeronFae]

You're trying to run on CPU, which is causing a Runtime error due to data type mismatch. I fixed it in my fork since I was having the exact same issue running the code on CPU only. I already placed a merge request to the main branch so they can fix the dtype error, since CPUs don't play nice with halved Float16 instructions to double processing speed; CPU instruction sets default to Float32, so they don't know how to compute proper mathematical operations on Float16 pairs unless you specify the AVX operations yourself.