black format

projects/Qwen/configs/qwen_sft.py

@@ -33,8 +33,8 @@
 tokenization = OmegaConf.create()
 tokenization.make_vocab_size_divisible_by = 1
 tokenization.tokenizer = LazyCall(Qwen2Tokenizer)(
-    vocab_file=pretrained_model_path+"/vocab.json",
-    merges_file=pretrained_model_path+"/merges.txt",
+    vocab_file=pretrained_model_path + "/vocab.json",
+    merges_file=pretrained_model_path + "/merges.txt",
 )
 
 

projects/Qwen/pipeline.py

@@ -74,6 +74,7 @@ def _parse_parameters(self, **pipeline_parameters):
     def preprocess(self, inputs, **kwargs) -> dict:
         # tokenizer encoderW
         import oneflow as flow
+
         inputs = flow.tensor(self.tokenizer.encode(inputs, add_bos=True, padding=True))
 
         inputs = {

projects/Qwen/utils/prepare_alpaca.py

@@ -111,12 +111,10 @@ def prepare_sample(example: dict, tokenizer, max_length: int) -> dict:
     """
     full_prompt = generate_prompt(example)
     full_prompt_and_response = full_prompt + example["output"]
-    
+
     prompt = tokenizer.encode(full_prompt, device="cpu")
     prompt = flow.tensor(prompt, dtype=flow.int, device="cpu")
-    example = tokenizer.encode(
-        full_prompt_and_response, device="cpu"
-    )
+    example = tokenizer.encode(full_prompt_and_response, device="cpu")
     example = flow.tensor(example, dtype=flow.int, device="cpu")
 
     padding = max_length - example.shape[0]