Merge branch 'master' of https://github.com/ZiyiXia/FlagEmbedding

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 @@ class BaseLLMEmbedder(AbsEmbedder):
         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 @@ class BaseLLMEmbedder(AbsEmbedder):
             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

@@ -54,6 +54,8 @@ 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`.
+        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`.
     
     Attributes:
         DEFAULT_POOLING_METHOD: The default pooling method when running the model.
@@ -77,8 +79,6 @@ class ICLLLMEmbedder(AbsEmbedder):
         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,
     ):
@@ -92,10 +92,8 @@ class ICLLLMEmbedder(AbsEmbedder):
             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 @@ class BaseEmbedder(AbsEmbedder):
         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 @@ class BaseEmbedder(AbsEmbedder):
             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 @@ class BaseEmbedder(AbsEmbedder):
         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 @@ class M3Embedder(AbsEmbedder):
         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 @@ class M3Embedder(AbsEmbedder):
             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,

research/BGE_M3/modeling.py

@@ -221,7 +221,7 @@ class BGEM3Model(nn.Module):
             if teacher_scores is not None:
                 # print("Use soft-label distillation...")
                 teacher_targets = F.softmax(teacher_scores, dim=-1)  # B N
-                group_size = p_sparse_vecs.size(0) // q_sparse_vecs.size(0)
+                group_size = p_dense_vecs.size(0) // q_dense_vecs.size(0)
 
                 # dense loss
                 dense_scores = self.dense_score(q_dense_vecs, p_dense_vecs)  # B, B * N

research/visual_bge/setup.py

@@ -0,0 +1,18 @@
+from setuptools import setup, find_packages
+
+setup(
+    name="visual_bge",
+    version="0.1.0",
+    description='visual_bge',
+    long_description="./README.md",
+    long_description_content_type="text/markdown",
+    url='https://github.com/FlagOpen/FlagEmbedding/tree/master/research/visual_bge',
+    packages=find_packages(),
+    install_requires=[
+        'torchvision',
+        'timm',
+        'einops',
+        'ftfy'
+    ],
+    python_requires='>=3.6',
+)

research/visual_bge/visual_bge/eva_clip/modified_resnet.py

@@ -4,7 +4,7 @@ import torch
 from torch import nn
 from torch.nn import functional as F
 
-from FlagEmbedding.visual.eva_clip.utils import freeze_batch_norm_2d
+from visual_bge.eva_clip.utils import freeze_batch_norm_2d
 
 
 class Bottleneck(nn.Module):