thenlper tomaarsen HF staff commited on
Commit
7fc0678
1 Parent(s): f7d567e

Fix broken SentenceTransformer snippet; format code with Python format (#11)

Browse files

- Fix broken SentenceTransformer snippet; format code with Python format (4f98c8de229b79178923ab4b65fa661c1dbf7b9e)


Co-authored-by: Tom Aarsen <tomaarsen@users.noreply.huggingface.co>

Files changed (1) hide show
  1. README.md +6 -14
README.md CHANGED
@@ -4660,7 +4660,7 @@ refer to [enable-unpadding-and-xformers](https://huggingface.co/Alibaba-NLP/new-
4660
 
4661
 
4662
  ### Get Dense Embeddings with Transformers
4663
- ```
4664
  # Requires transformers>=4.36.0
4665
 
4666
  import torch.nn.functional as F
@@ -4693,12 +4693,10 @@ print(scores.tolist())
4693
  ```
4694
 
4695
  ### Use with sentence-transformers
4696
- ```
4697
  # Requires sentences-transformers>=3.0.0
4698
 
4699
  from sentence_transformers import SentenceTransformer
4700
- from sentence_transformers.util import cos_sim
4701
- import numpy as np
4702
 
4703
  input_texts = [
4704
  "what is the capital of China?",
@@ -4708,24 +4706,18 @@ input_texts = [
4708
  ]
4709
 
4710
  model_name_or_path="Alibaba-NLP/gte-multilingual-base"
4711
- model = SentenceTransformer(', trust_remote_code=True)
4712
- embeddings = model.encode(input_texts) # embeddings.shape (4, 768)
4713
-
4714
- # normalized embeddings
4715
- norms = np.linalg.norm(embeddings, ord=2, axis=1, keepdims=True)
4716
- norms[norms == 0] = 1
4717
- embeddings = embeddings / norms
4718
 
4719
  # sim scores
4720
- scores = (embeddings[:1] @ embeddings[1:].T)
4721
 
4722
  print(scores.tolist())
4723
  # [[0.301699697971344, 0.7503870129585266, 0.32030850648880005]]
4724
-
4725
  ```
4726
 
4727
  ### Use with custom code to get dense embeddigns and sparse token weights
4728
- ```
4729
  # You can find the script gte_embedding.py in https://huggingface.co/Alibaba-NLP/gte-multilingual-base/blob/main/scripts/gte_embedding.py
4730
 
4731
  from gte_embedding import GTEEmbeddidng
 
4660
 
4661
 
4662
  ### Get Dense Embeddings with Transformers
4663
+ ```python
4664
  # Requires transformers>=4.36.0
4665
 
4666
  import torch.nn.functional as F
 
4693
  ```
4694
 
4695
  ### Use with sentence-transformers
4696
+ ```python
4697
  # Requires sentences-transformers>=3.0.0
4698
 
4699
  from sentence_transformers import SentenceTransformer
 
 
4700
 
4701
  input_texts = [
4702
  "what is the capital of China?",
 
4706
  ]
4707
 
4708
  model_name_or_path="Alibaba-NLP/gte-multilingual-base"
4709
+ model = SentenceTransformer(model_name_or_path, trust_remote_code=True)
4710
+ embeddings = model.encode(input_texts, normalize_embeddings=True) # embeddings.shape (4, 768)
 
 
 
 
 
4711
 
4712
  # sim scores
4713
+ scores = model.similarity(embeddings[:1], embeddings[1:])
4714
 
4715
  print(scores.tolist())
4716
  # [[0.301699697971344, 0.7503870129585266, 0.32030850648880005]]
 
4717
  ```
4718
 
4719
  ### Use with custom code to get dense embeddigns and sparse token weights
4720
+ ```python
4721
  # You can find the script gte_embedding.py in https://huggingface.co/Alibaba-NLP/gte-multilingual-base/blob/main/scripts/gte_embedding.py
4722
 
4723
  from gte_embedding import GTEEmbeddidng