MLX LM LoRA Fine Tune.ipynb

Downloaded and ran your notebook:

When calling the train() function from mlx_lm.tuner.trainer, it passes a tokenizer argument, which led to the following error:

TypeError: train() got an unexpected keyword argument 'tokenizer'

Checked the function signature using help(train):

Help on function train in module mlx_lm.tuner.trainer:

train(
model,
optimizer,
train_dataset,
val_dataset,
args: mlx_lm.tuner.trainer.TrainingArgs = TrainingArgs(...),
loss: callable = default_loss,
iterate_batches: callable = iterate_batches,
training_callback: TrainingCallback = None
)

It states that tokenizer is not a valid argument...but if you remove it then it cascades into another error.

remove the tokenizer from there and use train_dataset=CacheDataset(ds_train),
val_dataset=CacheDataset(ds_valid),