[BUG] Outputs from "cellpose.train.train_seg" are unclear

[braymp]

Describe the bug
Several sources indicate that there should be three outputs from train_seg, i.e., model_path, train_losses and test_losses:

• The definition for cellpose.train.train_seg describes the three outputs above. The output from this function indicates the same.
• If calling cellpose for training from the cmd line, the associated code yields just the first element of the output. The GUI code is similar.
• The test for train_seg obtains the path as the first element of the output.

That said, using Cellpose 3.1.0 to call cellpose.train.train_seg directly from the python cmd line with the three outputs specified yields the error:

Traceback (most recent call last):
File "", line 1, in
TypeError: cannot unpack non-iterable PosixPath object

Calling with just one output succeeds with path to the new model as the only output.

The API documentation for the function is ambiguous in that it says three outputs are returned but only describes the type for the path. The GitHub training documentation also indicates that three outputs are possible.

Please tell me if there's something I'm missing.

To Reproduce
I am calling cellpose.train.train_seg from the python cmd line to retrain the cyto3 Cellpose model:

model = cellpose.models.CellposeModel(gpu=True, model_type='cyto3', pretrained_model='cyto3')
model_path, train_losses, test_losses = cellpose.train.train_seg(net=model.net,
                         train_data=train_data,
                         train_labels=train_labels,
                         test_data=test_data,
                         test_labels=test_labels,
                         channels=channels,
                         model_name="my_new_model",
                         save_path="./test_path/",
                         )

Screenshots