remove conflicts

Former-commit-id: f8b637eb76cba7ec229e2978068805ad1cca8adb

src/llmtuner/hparams/data_args.py

@@ -26,11 +26,11 @@ class DataArguments:
     )
     cutoff_len: int = field(
         default=1024,
-        metadata={"help": "The cutoff length of the model inputs after tokenization."},
+        metadata={"help": "The cutoff length of the tokenized inputs in the dataset."},
     )
     reserved_label_len: int = field(
         default=1,
-        metadata={"help": "The minimum cutoff length reserved for label after tokenization."},
+        metadata={"help": "The minimum cutoff length reserved for the tokenized labels in the dataset."},
     )
     train_on_prompt: bool = field(
         default=False,

src/llmtuner/hparams/generating_args.py

@@ -31,11 +31,11 @@ class GeneratingArguments:
         metadata={"help": "Number of beams for beam search. 1 means no beam search."},
     )
     max_length: int = field(
-        default=512,
+        default=1024,
         metadata={"help": "The maximum length the generated tokens can have. It can be overridden by max_new_tokens."},
     )
     max_new_tokens: int = field(
-        default=512,
+        default=1024,
         metadata={"help": "The maximum numbers of tokens to generate, ignoring the number of tokens in the prompt."},
     )
     repetition_penalty: float = field(

src/llmtuner/hparams/model_args.py

@@ -22,7 +22,7 @@ class ModelArguments:
         metadata={"help": "Where to store the pre-trained models downloaded from huggingface.co or modelscope.cn."},
     )
     use_fast_tokenizer: bool = field(
-        default=False,
+        default=True,
         metadata={"help": "Whether or not to use one of the fast tokenizer (backed by the tokenizers library)."},
     )
     resize_vocab: bool = field(
@@ -33,6 +33,10 @@ class ModelArguments:
         default=False,
         metadata={"help": "Whether or not the special tokens should be split during the tokenization process."},
     )
+    new_special_tokens: Optional[str] = field(
+        default=None,
+        metadata={"help": "Special tokens to be added into the tokenizer."},
+    )
     model_revision: str = field(
         default="main",
         metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
@@ -61,9 +65,9 @@ class ModelArguments:
         default=None,
         metadata={"help": "Which scaling strategy should be adopted for the RoPE embeddings."},
     )
-    flash_attn: bool = field(
-        default=False,
-        metadata={"help": "Enable FlashAttention for faster training."},
+    flash_attn: Literal["off", "sdpa", "fa2", "auto"] = field(
+        default="auto",
+        metadata={"help": "Enable FlashAttention for faster training and inference."},
     )
     shift_attn: bool = field(
         default=False,
@@ -177,6 +181,9 @@ class ModelArguments:
         if self.adapter_name_or_path is not None:  # support merging multiple lora weights
             self.adapter_name_or_path = [path.strip() for path in self.adapter_name_or_path.split(",")]
 
+        if self.new_special_tokens is not None:  # support multiple special tokens
+            self.new_special_tokens = [token.strip() for token in self.new_special_tokens.split(",")]
+
         assert self.quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."
         assert self.export_quantization_bit in [None, 8, 4, 3, 2], "We only accept 2/3/4/8-bit quantization."
 

src/llmtuner/hparams/parser.py

@@ -67,6 +67,9 @@ def _verify_model_args(model_args: "ModelArguments", finetuning_args: "Finetunin
         if finetuning_args.finetuning_type != "lora":
             raise ValueError("Quantization is only compatible with the LoRA method.")
 
+        if model_args.resize_vocab:
+            raise ValueError("Cannot resize embedding layers of a quantized model.")
+
         if model_args.adapter_name_or_path is not None and finetuning_args.create_new_adapter:
             raise ValueError("Cannot create new adapter upon a quantized model.")
 
@@ -199,10 +202,11 @@ def get_train_args(args: Optional[Dict[str, Any]] = None) -> _TRAIN_CLS:
     if (
         training_args.do_train
         and finetuning_args.finetuning_type == "lora"
+        and model_args.quantization_bit is None
         and model_args.resize_vocab
         and finetuning_args.additional_target is None
     ):
-        logger.warning("Add token embeddings to `additional_target` to make the added tokens trainable.")
+        logger.warning("Remember to add embedding layers to `additional_target` to make the added tokens trainable.")
 
     if training_args.do_train and model_args.quantization_bit is not None and (not model_args.upcast_layernorm):
         logger.warning("We recommend enable `upcast_layernorm` in quantized training.")