Update README.md
Browse files
README.md
CHANGED
|
@@ -85,6 +85,46 @@ passage_embeddings = model.encode(
|
|
| 85 |
```
|
| 86 |
</details>
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
## Training & Evaluation
|
| 89 |
|
| 90 |
Please refer to our technical report of jina-embeddings-c1 for training details and benchmarks.
|
|
|
|
| 85 |
```
|
| 86 |
</details>
|
| 87 |
|
| 88 |
+
<details>
|
| 89 |
+
<summary>via <a href="https://sbert.net/">sentence-transformers</a></summary>
|
| 90 |
+
|
| 91 |
+
```python
|
| 92 |
+
# !pip install sentence_transformers>=5.0.0 torch>=2.7.1
|
| 93 |
+
|
| 94 |
+
import torch
|
| 95 |
+
from sentence_transformers import SentenceTransformer
|
| 96 |
+
|
| 97 |
+
# Load the model
|
| 98 |
+
model = SentenceTransformer(
|
| 99 |
+
"jinaai/jina-embeddings-c1-0.5B",
|
| 100 |
+
model_kwargs={
|
| 101 |
+
"torch_dtype": torch.bfloat16,
|
| 102 |
+
"attn_implementation": "flash_attention_2",
|
| 103 |
+
"device_map": "auto"
|
| 104 |
+
}
|
| 105 |
+
)
|
| 106 |
+
|
| 107 |
+
# The queries and documents to embed
|
| 108 |
+
queries = [
|
| 109 |
+
"print hello world in python",
|
| 110 |
+
"initialize array of 5 zeros in c++"
|
| 111 |
+
]
|
| 112 |
+
documents = [
|
| 113 |
+
"print('Hello World!')",
|
| 114 |
+
"int arr[5] = {0, 0, 0, 0, 0};"
|
| 115 |
+
]
|
| 116 |
+
|
| 117 |
+
query_embeddings = model.encode(queries, prompt_name="nl2code_query")
|
| 118 |
+
document_embeddings = model.encode(documents, prompt_name="nl2code_document")
|
| 119 |
+
|
| 120 |
+
# Compute the (cosine) similarity between the query and document embeddings
|
| 121 |
+
similarity = model.similarity(query_embeddings, document_embeddings)
|
| 122 |
+
print(similarity)
|
| 123 |
+
# tensor([[0.8157, 0.1222],
|
| 124 |
+
# [0.1201, 0.5500]])
|
| 125 |
+
```
|
| 126 |
+
</details>
|
| 127 |
+
|
| 128 |
## Training & Evaluation
|
| 129 |
|
| 130 |
Please refer to our technical report of jina-embeddings-c1 for training details and benchmarks.
|