Commit
•
48148d2
1
Parent(s):
17673ab
Test.
Browse files- README.md +76 -0
- pytorch_model.bin +1 -1
README.md
ADDED
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
tags:
|
3 |
+
- sentence-transformers
|
4 |
+
---
|
5 |
+
|
6 |
+
# TODO: Name of Model
|
7 |
+
|
8 |
+
TODO: Description
|
9 |
+
|
10 |
+
## Model Description
|
11 |
+
TODO: Add relevant content
|
12 |
+
|
13 |
+
(0) Base Transformer Type: RobertaModel
|
14 |
+
(1) Pooling mean
|
15 |
+
|
16 |
+
## Usage (Sentence-Transformers)
|
17 |
+
|
18 |
+
Using this model becomes more convenient when you have [sentence-transformers](https://github.com/UKPLab/sentence-transformers) installed:
|
19 |
+
|
20 |
+
```pip install -U sentence-transformers```
|
21 |
+
|
22 |
+
Then you can use the model like this:
|
23 |
+
|
24 |
+
```
|
25 |
+
from sentence_transformers import SentenceTransformer
|
26 |
+
sentences = ["This is an example sentence"]
|
27 |
+
|
28 |
+
model = SentenceTransformer(TODO)
|
29 |
+
embeddings = model.encode(sentences)
|
30 |
+
print(embeddings)
|
31 |
+
```
|
32 |
+
|
33 |
+
|
34 |
+
## Usage (HuggingFace Transformers)
|
35 |
+
|
36 |
+
```
|
37 |
+
from transformers import AutoTokenizer, AutoModel
|
38 |
+
import torch
|
39 |
+
|
40 |
+
# The next step is optional if you want your own pooling function.
|
41 |
+
# Max Pooling - Take the max value over time for every dimension.
|
42 |
+
def max_pooling(model_output, attention_mask):
|
43 |
+
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
|
44 |
+
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
|
45 |
+
token_embeddings[input_mask_expanded == 0] = -1e9 # Set padding tokens to large negative value
|
46 |
+
max_over_time = torch.max(token_embeddings, 1)[0]
|
47 |
+
return max_over_time
|
48 |
+
|
49 |
+
#Sentences we want sentence embeddings for
|
50 |
+
sentences = ['This is an example sentence']
|
51 |
+
|
52 |
+
#Load model from HuggingFace Hub
|
53 |
+
tokenizer = AutoTokenizer.from_pretrained(TODO)
|
54 |
+
model = AutoModel.from_pretrained(TODO
|
55 |
+
|
56 |
+
#Tokenize sentences
|
57 |
+
encoded_input = tokenizer(sentences, padding=True, truncation=True, max_length=128, return_tensors='pt'))
|
58 |
+
|
59 |
+
#Compute token embeddings
|
60 |
+
with torch.no_grad():
|
61 |
+
model_output = model(**encoded_input)
|
62 |
+
|
63 |
+
#Perform pooling. In this case, max pooling.
|
64 |
+
sentence_embeddings = max_pooling(model_output, encoded_input['attention_mask'])
|
65 |
+
|
66 |
+
print("Sentence embeddings:")
|
67 |
+
print(sentence_embeddings)
|
68 |
+
```
|
69 |
+
|
70 |
+
|
71 |
+
|
72 |
+
## TODO: Training Procedure
|
73 |
+
|
74 |
+
## TODO: Evaluation Results
|
75 |
+
|
76 |
+
## TODO: Citing & Authors
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 328519167
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:02be9d4693ea9f8c4ed5beeefd366ade0c27ccf6775c4b6f4ccd1628041b1262
|
3 |
size 328519167
|