Training fails in weird way if you don't provide validation data

[dadmobile]

It looks like as part of the update to use the mlx_lx, we not longer halt if the training dataset doesn't contain validation data. As a result, no adapter is created and that ends up being what comes up as an error to the user.

Eg.

Original error earlier in training:

ValueError: Validation set not found or empty. Must provide validation set for fine-tuning. Loading datasets Finished training. Now fusing the adaptor with the model.

And then later:

FileNotFoundError: [Errno 2] No such file or directory: '/Users/hudsongouge/.transformerlab/workspace/plugins/mlx_lora_trainer/adapter_config.json' Return code: 1 Fusing model with adaptor failed: 1