fix val.py bug

[regainOWO]

I found create_dataloader function param names variable type is list, but in val.py line 161

names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)}

now names is a dict, this will make error if you didn't train A dataset before, if you train A dataset before, it will create label cache and create_dataloader will load label cache, no errors will occur.
maybe it in line 167, it can change to

dataloader = create_dataloader(data[task], imgsz, batch_size, stride, list(names.values()), single_cls, pad=pad, rect=pt,
                                       workers=workers, prefix=colorstr(f'{task}: '))[0]