Update README.md
Browse files
README.md
CHANGED
@@ -1130,7 +1130,7 @@ def mean_pooling(model_output, attention_mask):
|
|
1130 |
sentences = ['How is the weather today?', '今天天气怎么样?']
|
1131 |
|
1132 |
tokenizer = AutoTokenizer.from_pretrained('jinaai/jina-embeddings-v2-base-zh')
|
1133 |
-
model = AutoModel.from_pretrained('jinaai/jina-embeddings-v2-base-zh', trust_remote_code=True)
|
1134 |
|
1135 |
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
|
1136 |
|
@@ -1148,11 +1148,12 @@ You can use Jina Embedding models directly from transformers package.
|
|
1148 |
|
1149 |
```python
|
1150 |
!pip install transformers
|
|
|
1151 |
from transformers import AutoModel
|
1152 |
from numpy.linalg import norm
|
1153 |
|
1154 |
cos_sim = lambda a,b: (a @ b.T) / (norm(a)*norm(b))
|
1155 |
-
model = AutoModel.from_pretrained('jinaai/jina-embeddings-v2-base-zh', trust_remote_code=True)
|
1156 |
embeddings = model.encode(['How is the weather today?', '今天天气怎么样?'])
|
1157 |
print(cos_sim(embeddings[0], embeddings[1]))
|
1158 |
```
|
|
|
1130 |
sentences = ['How is the weather today?', '今天天气怎么样?']
|
1131 |
|
1132 |
tokenizer = AutoTokenizer.from_pretrained('jinaai/jina-embeddings-v2-base-zh')
|
1133 |
+
model = AutoModel.from_pretrained('jinaai/jina-embeddings-v2-base-zh', trust_remote_code=True, torch_dtype=torch.bfloat16)
|
1134 |
|
1135 |
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
|
1136 |
|
|
|
1148 |
|
1149 |
```python
|
1150 |
!pip install transformers
|
1151 |
+
import torch
|
1152 |
from transformers import AutoModel
|
1153 |
from numpy.linalg import norm
|
1154 |
|
1155 |
cos_sim = lambda a,b: (a @ b.T) / (norm(a)*norm(b))
|
1156 |
+
model = AutoModel.from_pretrained('jinaai/jina-embeddings-v2-base-zh', trust_remote_code=True, torch_dtype=torch.bfloat16)
|
1157 |
embeddings = model.encode(['How is the weather today?', '今天天气怎么样?'])
|
1158 |
print(cos_sim(embeddings[0], embeddings[1]))
|
1159 |
```
|