Default dataset to cleaned alpaca dataset from HF (#202)

finetune.py

@@ -26,7 +26,7 @@ from transformers import LlamaForCausalLM, LlamaTokenizer  # noqa: F402
 def train(
     # model/data params
     base_model: str = "",  # the only required argument
-    data_path: str = "./alpaca_data_cleaned.json",
+    data_path: str = "yahma/alpaca-cleaned",
     output_dir: str = "./lora-alpaca",
     # training hyperparams
     batch_size: int = 128,