Update README.md
Browse files
README.md
CHANGED
|
@@ -1789,31 +1789,14 @@ print(finetuner.cos_sim(embeddings[0], embeddings[1]))
|
|
| 1789 |
Use directly with Huggingface Transformers:
|
| 1790 |
|
| 1791 |
```python
|
| 1792 |
-
import
|
| 1793 |
-
from
|
| 1794 |
-
|
| 1795 |
-
|
| 1796 |
-
def mean_pooling(model_output, attention_mask):
|
| 1797 |
-
token_embeddings = model_output[0]
|
| 1798 |
-
input_mask_expanded = (
|
| 1799 |
-
attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
|
| 1800 |
-
)
|
| 1801 |
-
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(
|
| 1802 |
-
input_mask_expanded.sum(1), min=1e-9
|
| 1803 |
-
)
|
| 1804 |
|
| 1805 |
sentences = ['how is the weather today', 'What is the current weather like today?']
|
| 1806 |
|
| 1807 |
-
|
| 1808 |
-
|
| 1809 |
-
|
| 1810 |
-
|
| 1811 |
-
with torch.inference_mode():
|
| 1812 |
-
encoded_input = tokenizer(
|
| 1813 |
-
sentences, padding=True, truncation=True, return_tensors='pt'
|
| 1814 |
-
)
|
| 1815 |
-
model_output = model.encoder(**encoded_input)
|
| 1816 |
-
embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
|
| 1817 |
```
|
| 1818 |
|
| 1819 |
## Fine-tuning
|
|
|
|
| 1789 |
Use directly with Huggingface Transformers:
|
| 1790 |
|
| 1791 |
```python
|
| 1792 |
+
from sentence_transformers import SentenceTransformer
|
| 1793 |
+
from sentence_transformers.util import cos_sim
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1794 |
|
| 1795 |
sentences = ['how is the weather today', 'What is the current weather like today?']
|
| 1796 |
|
| 1797 |
+
model = SentenceTransformer('jinaai/jina-embedding-s-en-v1')
|
| 1798 |
+
embeddings = model.encode(sentences)
|
| 1799 |
+
print(cos_sim(embeddings[0], embeddings[1]))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1800 |
```
|
| 1801 |
|
| 1802 |
## Fine-tuning
|