update

Author: minmingzhu

llm_on_ray/finetune/finetune.yaml

@@ -13,7 +13,8 @@ General:
     lora_dropout: 0.1
   enable_gradient_checkpointing: false
 Dataset:
-  train_file: examples/data/sample_finetune_data_small.jsonl
+  type: "SlimOrca"
+  train_file: Open-Orca/SlimOrca
   group: true
   max_length: 512
   block_size: 512

llm_on_ray/finetune/finetune_config.py

@@ -80,6 +80,7 @@ class General(BaseModel):
 
 
 class Dataset(BaseModel):
+    type: str = "chat"
     train_file: str
     validation_file: Optional[str]
     validation_split_percentage: int