File size: 8,773 Bytes
7b03a43 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 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 |
---
language:
- zh
- en
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- transformers
pipeline_tag: sentence-similarity
library_name: sentence-transformers
license: apache-2.0
---
<h1 align="center">FlagEmbedding</h1>
For more details please refer to our Github: [FlagEmbedding](https://github.com/FlagOpen/FlagEmbedding).
**BGE-Code-v1** is an LLM-based code embedding model that supports code retrieval, text retrieval, and multilingual retrieval. It primarily demonstrates the following capabilities:
- Superior Code Retrieval Performance: The model demonstrates exceptional code retrieval capabilities, supporting natural language queries in both English and Chinese, as well as 20 programming languages.
- Robust Text Retrieval Capabilities: The model maintains strong text retrieval capabilities comparable to text embedding models of similar scale.
- Extensive Multilingual Support: BGE-Code-v1 offers comprehensive multilingual retrieval capabilities, excelling in languages such as English, Chinese, Japanese, French, and more.
## Usage
### Using FlagEmbedding
```
git clone https://github.com/FlagOpen/FlagEmbedding.git
cd FlagEmbedding
pip install -e .
```
```python
from FlagEmbedding import FlagLLMModel
queries = [
"Delete the record with ID 4 from the 'Staff' table.",
'Delete all records in the "Livestock" table where age is greater than 5'
]
documents = [
"DELETE FROM Staff WHERE StaffID = 4;",
"DELETE FROM Livestock WHERE age > 5;"
]
model = FlagLLMModel('BAAI/BGE-Code-v1',
query_instruction_format="<instruct>{}\n<query>{}",
query_instruction_for_retrieval="Given a question in text, retrieve SQL queries that are appropriate responses to the question.",
trust_remote_code=True,
use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
embeddings_1 = model.encode_queries(queries)
embeddings_2 = model.encode_corpus(documents)
similarity = embeddings_1 @ embeddings_2.T
print(similarity)
```
By default, FlagLLMModel will use all available GPUs when encoding. Please set `os.environ["CUDA_VISIBLE_DEVICES"]` to select specific GPUs. You also can set `os.environ["CUDA_VISIBLE_DEVICES"]=""` to make all GPUs unavailable.
### Using Sentence Transformers
```python
from sentence_transformers import SentenceTransformer
import torch
# Load the model, optionally in float16 precision for faster inference
model = SentenceTransformer("BAAI/bge-code-v1", model_kwargs={"torch_dtype": torch.float16, "trust_remote_code": True}, tokenizer_kwargs={"trust_remote_code": True})
# Prepare a prompt given an instruction
instruction = 'Given a question in text, retrieve SQL queries that are appropriate responses to the question.'
prompt = f'<instruct>{instruction}\n<query>'
# Prepare queries and documents
queries = [
"Delete the record with ID 4 from the 'Staff' table.",
'Delete all records in the "Livestock" table where age is greater than 5'
]
documents = [
"DELETE FROM Staff WHERE StaffID = 4;",
"DELETE FROM Livestock WHERE age > 5;"
]
# Compute the query and document embeddings
query_embeddings = model.encode(queries, prompt=prompt)
document_embeddings = model.encode(documents)
# Compute the cosine similarity between the query and document embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
```
### Using HuggingFace Transformers
```python
import torch
import torch.nn.functional as F
from torch import Tensor
from transformers import AutoTokenizer, AutoModel
def last_token_pool(last_hidden_states: Tensor,
attention_mask: Tensor) -> Tensor:
left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0])
if left_padding:
return last_hidden_states[:, -1]
else:
sequence_lengths = attention_mask.sum(dim=1) - 1
batch_size = last_hidden_states.shape[0]
return last_hidden_states[torch.arange(batch_size, device=last_hidden_states.device), sequence_lengths]
def get_detailed_instruct(task_description: str, query: str) -> str:
return f'<instruct>{task_description}\n<query>{query}'
instruction = 'Given a question in text, retrieve SQL queries that are appropriate responses to the question.'
queries = [
"Delete the record with ID 4 from the 'Staff' table.",
'Delete all records in the "Livestock" table where age is greater than 5'
]
documents = [
"DELETE FROM Staff WHERE StaffID = 4;",
"DELETE FROM Livestock WHERE age > 5;"
]
input_texts = queries + documents
tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-code-v1', trust_remote_code=True)
model = AutoModel.from_pretrained('BAAI/bge-code-v1', trust_remote_code=True)
model.eval()
max_length = 4096
# Tokenize the input texts
batch_dict = tokenizer(input_texts, max_length=max_length, padding=True, truncation=True, return_tensors='pt', pad_to_multiple_of=8)
with torch.no_grad():
outputs = model(**batch_dict)
embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask'])
# normalize embeddings
embeddings = F.normalize(embeddings, p=2, dim=1)
scores = (embeddings[:2] @ embeddings[2:].T) * 100
print(scores.tolist())
```
## Evaluation
**BGE-Code-v1** achieves state-of-the-art performance on both the CoIR and CodeRAG benchmarks.
- CoIR
| | CodeXEmbed-2B | CodeXEmbed-7B | Voyage-Code-002 | Voyage-Code-003 | BGE-Code-v1 |
|---------------------------------------|---------------|---------------|-----------------|-----------------|-----------|
| Apps | 76.86 | 85.38 | 26.52 | 93.62 | 98.08 |
| CosQA | 40.47 | 42.47 | 29.79 | 34.45 | 46.72 |
| Text2SQL | 78.42 | 78.94 | 69.26 | 62.87 | 64.35 |
| CSN | 87.87 | 89.67 | 81.79 | 89.35 | 89.53 |
| CSN-CCR | 97.66 | 97.95 | 73.45 | 90.05 | 98.30 |
| CodeTrans-Contest | 90.30 | 94.45 | 72.77 | 94.96 | 94.38 |
| CodeTrans-DL | 38.57 | 40.46 | 27.48 | 38.57 | 46.13 |
| StackOverFlow-QA | 94.47 | 96.33 | 67.68 | 97.17 | 95.35 |
| CodeFeedBack-ST | 86.36 | 87.53 | 65.35 | 90.67 | 90.56 |
| CodeFeedBack-MT | 65.51 | 68.83 | 28.74 | 93.58 | 94.38 |
| AVG | 75.65 | 78.20 | 56.26 | 78.53 | 81.77 |
- CodedRAG
| | HummanEval | MBPP | DS-1000 | ODEX | RepoEval | SWE-bench-Lite | AVG |
| --------------- | ---------- | ---- | ------- | ---- | -------- | -------------- | ---- |
| SFR | 100.0 | 99.0 | 19.3 | 37.1 | 83.8 | 62.7 | 67.0 |
| Jina-v2-code | 100.0 | 97.7 | 26.2 | 19.9 | 90.5 | 58.3 | 65.4 |
| CodeXEmbed-2B | 100.0 | 97.4 | 25.4 | 23.9 | 88.7 | 52.4 | 64.6 |
| Voyage-Code-002 | 100.0 | 99.0 | 33.1 | 26.6 | 94.3 | 29.1 | 63.7 |
| Voyage-Code-003 | 100.0 | 99.6 | 38.9 | 36.3 | 90.0 | 70.1 | 72.5 |
| BGE-Code-v1 | 100.0 | 99.2 | 40.9 | 36.1 | 93.1 | 67.4 | 72.8 |
## Citation
If you find this repository useful, please consider giving a star :star: and citation
```
@article{bge-llm,
title={Making text embedders few-shot learners},
author={Li, Chaofan and Qin, MingHao and Xiao, Shitao and Chen, Jianlyu and Luo, Kun and Shao, Yingxia and Lian, Defu and Liu, Zheng},
journal={arXiv preprint arXiv:2409.15700},
year={2024}
}
@misc{bge-m3,
title={BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation},
author={Jianlv Chen and Shitao Xiao and Peitian Zhang and Kun Luo and Defu Lian and Zheng Liu},
year={2024},
eprint={2402.03216},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
@misc{bge_embedding,
title={C-Pack: Packaged Resources To Advance General Chinese Embedding},
author={Shitao Xiao and Zheng Liu and Peitian Zhang and Niklas Muennighoff},
year={2023},
eprint={2309.07597},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` |