Some improvements to LoRA (#528)

* set cache_limit

* remove set cache_limit

* cleanup

* add gradient checkpointing

* fix sort

* mokey patch call for checkpoint

* fix example config

llms/mlx_lm/examples/lora_config.yaml

@@ -1,6 +1,5 @@
 # The path to the local model directory or Hugging Face repo.
 model: "mlx_model"
-
 # Whether or not to train (boolean)
 train: true
 
@@ -49,6 +48,9 @@ test_batches: 500
 # Maximum sequence length.
 max_seq_length: 2048
 
+# Use gradient checkpointing to reduce memory use.
+grad_checkpoint: false
+
 # LoRA parameters can only be specified in a config file
 lora_parameters:
   # The layer keys to apply LoRA to.