23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252 | class OpenVINOEmbedding(BaseEmbedding):
model_id_or_path: str = Field(description="Huggingface model id or local path.")
max_length: int = Field(description="Maximum length of input.")
pooling: str = Field(description="Pooling strategy. One of ['cls', 'mean'].")
normalize: bool = Field(default=True, description="Normalize embeddings or not.")
query_instruction: Optional[str] = Field(
description="Instruction to prepend to query text."
)
text_instruction: Optional[str] = Field(
description="Instruction to prepend to text."
)
cache_folder: Optional[str] = Field(
description="Cache folder for huggingface files.", default=None
)
_model: Any = PrivateAttr()
_tokenizer: Any = PrivateAttr()
_device: Any = PrivateAttr()
def __init__(
self,
model_id_or_path: str = "BAAI/bge-m3",
pooling: str = "cls",
max_length: Optional[int] = None,
normalize: bool = True,
query_instruction: Optional[str] = None,
text_instruction: Optional[str] = None,
model: Optional[Any] = None,
tokenizer: Optional[Any] = None,
embed_batch_size: int = DEFAULT_EMBED_BATCH_SIZE,
callback_manager: Optional[CallbackManager] = None,
model_kwargs: Dict[str, Any] = {},
device: Optional[str] = "auto",
):
try:
from huggingface_hub import HfApi
except ImportError as e:
raise ValueError(
"Could not import huggingface_hub python package. "
"Please install it with: "
"`pip install -U huggingface_hub`."
) from e
def require_model_export(
model_id: str, revision: Any = None, subfolder: Any = None
) -> bool:
model_dir = Path(model_id)
if subfolder is not None:
model_dir = model_dir / subfolder
if model_dir.is_dir():
return (
not (model_dir / "openvino_model.xml").exists()
or not (model_dir / "openvino_model.bin").exists()
)
hf_api = HfApi()
try:
model_info = hf_api.model_info(model_id, revision=revision or "main")
normalized_subfolder = (
None if subfolder is None else Path(subfolder).as_posix()
)
model_files = [
file.rfilename
for file in model_info.siblings
if normalized_subfolder is None
or file.rfilename.startswith(normalized_subfolder)
]
ov_model_path = (
"openvino_model.xml"
if subfolder is None
else f"{normalized_subfolder}/openvino_model.xml"
)
return (
ov_model_path not in model_files
or ov_model_path.replace(".xml", ".bin") not in model_files
)
except Exception:
return True
if require_model_export(model_id_or_path):
# use remote model
model = model or OVModelForFeatureExtraction.from_pretrained(
model_id_or_path, export=True, device=device, **model_kwargs
)
else:
# use local model
model = model or OVModelForFeatureExtraction.from_pretrained(
model_id_or_path, device=device, **model_kwargs
)
tokenizer = tokenizer or AutoTokenizer.from_pretrained(model_id_or_path)
if max_length is None:
try:
max_length = int(model.config.max_position_embeddings)
except Exception:
raise ValueError(
"Unable to find max_length from model config. "
"Please provide max_length."
)
try:
max_length = min(max_length, int(tokenizer.model_max_length))
except Exception as exc:
print(f"An error occurred while retrieving tokenizer max length: {exc}")
if pooling not in ["cls", "mean"]:
raise ValueError(f"Pooling {pooling} not supported.")
super().__init__(
embed_batch_size=embed_batch_size,
callback_manager=callback_manager or CallbackManager([]),
model_id_or_path=model_id_or_path,
max_length=max_length,
pooling=pooling,
normalize=normalize,
query_instruction=query_instruction,
text_instruction=text_instruction,
)
self._device = device
self._model = model
self._tokenizer = tokenizer
@classmethod
def class_name(cls) -> str:
return "OpenVINOEmbedding"
@staticmethod
def create_and_save_openvino_model(
model_name_or_path: str,
output_path: str,
export_kwargs: Optional[dict] = None,
) -> None:
try:
from optimum.intel.openvino import OVModelForFeatureExtraction
from transformers import AutoTokenizer
except ImportError:
raise ImportError(
"OpenVINO Embedding requires transformers and optimum to be installed.\n"
"Please install transformers with "
"`pip install transformers optimum[openvino]`."
)
export_kwargs = export_kwargs or {}
model = OVModelForFeatureExtraction.from_pretrained(
model_name_or_path, export=True, compile=False, **export_kwargs
)
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
model.save_pretrained(output_path)
tokenizer.save_pretrained(output_path)
print(
f"Saved OpenVINO model to {output_path}. Use it with "
f"`embed_model = OpenVINOEmbedding(model_id_or_path='{output_path}')`."
)
def _mean_pooling(self, model_output: Any, attention_mask: Any) -> Any:
"""Mean Pooling - Take attention mask into account for correct averaging."""
import torch
# First element of model_output contains all token embeddings
token_embeddings = model_output[0]
input_mask_expanded = (
attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
)
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(
input_mask_expanded.sum(1), min=1e-9
)
def _cls_pooling(self, model_output: list) -> Any:
"""Use the CLS token as the pooling token."""
return model_output[0][:, 0]
def _embed(self, sentences: List[str]) -> List[List[float]]:
"""Embed sentences."""
length = self._model.request.inputs[0].get_partial_shape()[1]
if length.is_dynamic:
encoded_input = self._tokenizer(
sentences,
padding=True,
max_length=self.max_length,
truncation=True,
return_tensors="pt",
)
else:
encoded_input = self._tokenizer(
sentences,
padding="max_length",
max_length=length.get_length(),
truncation=True,
return_tensors="pt",
)
model_output = self._model(**encoded_input)
if self.pooling == "cls":
embeddings = self._cls_pooling(model_output)
else:
embeddings = self._mean_pooling(
model_output, encoded_input["attention_mask"]
)
if self.normalize:
import torch
embeddings = torch.nn.functional.normalize(embeddings, p=2, dim=1)
return embeddings.tolist()
def _get_query_embedding(self, query: str) -> List[float]:
"""Get query embedding."""
query = format_query(query, self.model_name, self.query_instruction)
return self._embed([query])[0]
async def _aget_query_embedding(self, query: str) -> List[float]:
"""Get query embedding async."""
return self._get_query_embedding(query)
async def _aget_text_embedding(self, text: str) -> List[float]:
"""Get text embedding async."""
return self._get_text_embedding(text)
def _get_text_embedding(self, text: str) -> List[float]:
"""Get text embedding."""
text = format_text(text, self.model_name, self.text_instruction)
return self._embed([text])[0]
def _get_text_embeddings(self, texts: List[str]) -> List[List[float]]:
"""Get text embeddings."""
texts = [
format_text(text, self.model_name, self.text_instruction) for text in texts
]
return self._embed(texts)
|
