Merge pull request #1189 from hanhainebula/master

delete useless parameters for embedder classes

Author: 545999961

FlagEmbedding/inference/embedder/decoder_only/base.py

@@ -47,8 +47,6 @@ class BaseLLMEmbedder(AbsEmbedder):
         batch_size (int, optional): Batch size for inference. Defaults to :data:`256`.
         query_max_length (int, optional): Maximum length for query. Defaults to :data:`512`.
         passage_max_length (int, optional): Maximum length for passage. Defaults to :data:`512`.
-        instruction (Optional[str], optional): Instruction for embedding with :attr:`instruction_format`. Defaults to :data:`None`.
-        instruction_format (str, optional): Instruction format when using :attr:`instruction`. Defaults to :data:`"{}{}"`.
         convert_to_numpy (bool, optional): If True, the output embedding will be a Numpy array. Otherwise, it will be a Torch Tensor. 
             Defaults to :data:`True`.
     
@@ -72,8 +70,6 @@ def __init__(
         batch_size: int = 256,
         query_max_length: int = 512,
         passage_max_length: int = 512,
-        instruction: Optional[str] = None,
-        instruction_format: str = "{}{}",
         convert_to_numpy: bool = True,
         **kwargs: Any,
     ):
@@ -87,8 +83,6 @@ def __init__(
             batch_size=batch_size,
             query_max_length=query_max_length,
             passage_max_length=passage_max_length,
-            instruction=instruction,
-            instruction_format=instruction_format,
             convert_to_numpy=convert_to_numpy,
             **kwargs
         )

FlagEmbedding/inference/embedder/decoder_only/icl.py

@@ -54,8 +54,6 @@ class ICLLLMEmbedder(AbsEmbedder):
         batch_size (int, optional): Batch size for inference. Defaults to :data:`256`.
         query_max_length (int, optional): Maximum length for query. Defaults to :data:`512`.
         passage_max_length (int, optional): Maximum length for passage. Defaults to :data:`512`.
-        instruction (Optional[str], optional): Instruction for embedding with :attr:`instruction_format`. Defaults to :data:`None`.
-        instruction_format (str, optional): Instruction format when using :attr:`instruction`. Defaults to :data:`"{}{}"`.
         convert_to_numpy (bool, optional): If True, the output embedding will be a Numpy array. Otherwise, it will be a Torch Tensor. 
             Defaults to :data:`True`.
     
@@ -81,8 +79,6 @@ def __init__(
         batch_size: int = 256,
         query_max_length: int = 512,
         passage_max_length: int = 512,
-        instruction: Optional[str] = None,
-        instruction_format: str = "{}{}",
         convert_to_numpy: bool = True,
         **kwargs: Any,
     ):
@@ -96,10 +92,8 @@ def __init__(
             batch_size=batch_size,
             query_max_length=query_max_length,
             passage_max_length=passage_max_length,
-            instruction=instruction,
-            instruction_format=instruction_format,
             convert_to_numpy=convert_to_numpy,
-            kwargs=kwargs
+            **kwargs
         )
 
         self.tokenizer = AutoTokenizer.from_pretrained(

FlagEmbedding/inference/embedder/encoder_only/base.py

@@ -28,8 +28,6 @@ class BaseEmbedder(AbsEmbedder):
         batch_size (int, optional): Batch size for inference. Defaults to :data:`256`.
         query_max_length (int, optional): Maximum length for query. Defaults to :data:`512`.
         passage_max_length (int, optional): Maximum length for passage. Defaults to :data:`512`.
-        instruction (Optional[str], optional): Instruction for embedding with :attr:`instruction_format`. Defaults to :data:`None`.
-        instruction_format (str, optional): Instruction format when using :attr:`instruction`. Defaults to :data:`"{}{}"`.
         convert_to_numpy (bool, optional): If True, the output embedding will be a Numpy array. Otherwise, it will be a Torch Tensor. 
             Defaults to :data:`True`.
     
@@ -55,8 +53,6 @@ def __init__(
         batch_size: int = 256,
         query_max_length: int = 512,
         passage_max_length: int = 512,
-        instruction: Optional[str] = None,
-        instruction_format: str = "{}{}",
         convert_to_numpy: bool = True,
         **kwargs: Any,
     ):
@@ -70,8 +66,6 @@ def __init__(
             batch_size=batch_size,
             query_max_length=query_max_length,
             passage_max_length=passage_max_length,
-            instruction=instruction,
-            instruction_format=instruction_format,
             convert_to_numpy=convert_to_numpy,
             **kwargs
         )
@@ -201,9 +195,6 @@ def encode_single_device(
         if device == "cpu": self.use_fp16 = False
         if self.use_fp16: self.model.half()
 
-        if device == "cpu": self.use_fp16 = False
-        if self.use_fp16: self.model.half()
-
         self.model.to(device)
         self.model.eval()
 

FlagEmbedding/inference/embedder/encoder_only/m3.py

@@ -38,8 +38,6 @@ class M3Embedder(AbsEmbedder):
         batch_size (int, optional): Batch size for inference. Defaults to :data:`256`.
         query_max_length (int, optional): Maximum length for query. Defaults to :data:`512`.
         passage_max_length (int, optional): Maximum length for passage. Defaults to :data:`512`.
-        instruction (Optional[str], optional): Instruction for embedding with :attr:`instruction_format`. Defaults to :data:`None`.
-        instruction_format (str, optional): Instruction format when using :attr:`instruction`. Defaults to :data:`"{}{}"`.
         return_dense (bool, optional): If true, will return the dense embedding. Defaults to :data:`True`.
         return_sparse (bool, optional): If true, will return the sparce embedding. Defaults to :data:`False`.
         return_colbert_vecs (bool, optional): If true, will return the colbert vectors. Defaults to :data:`False`.
@@ -66,8 +64,6 @@ def __init__(
         batch_size: int = 256,
         query_max_length: int = 512,
         passage_max_length: int = 512,
-        instruction: Optional[str] = None,
-        instruction_format: str = "{}{}",
         return_dense: bool = True,
         return_sparse: bool = False,
         return_colbert_vecs: bool = False,
@@ -83,8 +79,6 @@ def __init__(
             batch_size=batch_size,
             query_max_length=query_max_length,
             passage_max_length=passage_max_length,
-            instruction=instruction,
-            instruction_format=instruction_format,
             return_dense=return_dense,
             return_sparse=return_sparse,
             return_colbert_vecs=return_colbert_vecs,