[BUG]Error saving new trained model

[Ge0rges]

Describe the bug When I try to train a new model, an error occurs when saving.

To Reproduce Steps to reproduce the behavior: I followed the train new model instructions.

Run log Ran python -m cellpose got:

2025-10-08 11:34:05,762 [INFO] training with ['DSCF5955.JPG']
2025-10-08 11:34:05,766 [INFO] training new model starting at model cpsam
2025-10-08 11:34:05,768 [WARNING] model_type argument is not used in v4.0.1+. Ignoring this argument...
2025-10-08 11:34:05,811 [INFO] ** TORCH MPS version installed and working. **
2025-10-08 11:34:05,818 [INFO] >>>> using GPU (MPS)
2025-10-08 11:34:44,643 [INFO] >>>> loading model /Users/GeorgesKanaan/.cellpose/models/cpsam
GUI_INFO: name of new model: cpsam_train1
2025-10-08 11:38:08,910 [INFO] computing flows for labels
2025-10-08 11:39:42,853 [INFO] >>> computing diameters
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00,  7.80it/s]
2025-10-08 11:39:43,018 [INFO] >>> normalizing {'lowhigh': None, 'percentile': [1.0, 99.0], 'normalize': True, 'norm3D': True, 'sharpen_radius': 20.0, 'smooth_radius': 0.0, 'tile_norm_blocksize': 0.0, 'tile_norm_smooth3D': 1.0, 'invert': False}
2025-10-08 11:39:44,638 [INFO] >>> n_epochs=100, n_train=1, n_test=None
2025-10-08 11:39:44,638 [INFO] >>> AdamW, learning_rate=0.00001, weight_decay=0.10000
2025-10-08 11:39:44,673 [INFO] >>> saving model to /Users/GeorgesKanaan/Documents/Work/0_Projects/0_Memory/1_Cycling_Memory/Data/Scope/33S10_25ul/models/cpsam_train1
Traceback (most recent call last):
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/cellpose/gui/gui.py", line 1808, in new_model
    self.train_model(restore=restore, normalize_params=normalize_params)
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/cellpose/gui/gui.py", line 1825, in train_model
    self.new_model_path, train_losses = train.train_seg(
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/cellpose/train.py", line 468, in train_seg
    loss.backward()
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/torch/_tensor.py", line 648, in backward
    torch.autograd.backward(
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/torch/autograd/__init__.py", line 353, in backward
    _engine_run_backward(
  File "/Users/GeorgesKanaan/micromamba/envs/jupyter/lib/python3.10/site-packages/torch/autograd/graph.py", line 824, in _engine_run_backward
    return Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
RuntimeError: Expected scalar_type == ScalarType::Float || inputTensor.scalar_type() == ScalarType::Int || scalar_type == ScalarType::Bool to be true, but got false.  (Could this error message be improved?  If so, please report an enhancement request to PyTorch.)

[Jiadalee]

this seems an input datatype error. what image data did you use to train the model? .npy file or .jpg/.png file?