Update README.md
Browse filesAdd "how to use"
README.md
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
---
|
2 |
pipeline_tag: text-classification
|
|
|
3 |
language: en
|
4 |
tags:
|
5 |
- transformers
|
@@ -7,4 +8,37 @@ tags:
|
|
7 |
|
8 |
# Prompsit/paraphrase-bert-en
|
9 |
|
10 |
-
This model allows to evaluate paraphrases for a given phrase.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
pipeline_tag: text-classification
|
3 |
+
inference: false
|
4 |
language: en
|
5 |
tags:
|
6 |
- transformers
|
|
|
8 |
|
9 |
# Prompsit/paraphrase-bert-en
|
10 |
|
11 |
+
This model allows to evaluate paraphrases for a given phrase.
|
12 |
+
We have fine-tuned this model from pretrained "bert-base-uncased".
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
# How to usage
|
17 |
+
|
18 |
+
The model answer the following question: Is "phrase B" paraphrases of "phrase A".
|
19 |
+
Please note that we're considering phrases instead of sentences. Therefore, we must take into account that the model doesn't expect to find punctuation marks or long pieces of text.
|
20 |
+
|
21 |
+
Resulting probabilities correspond to classes:
|
22 |
+
* 0: Not a paraphrase
|
23 |
+
* 1: It's a paraphrase
|
24 |
+
|
25 |
+
|
26 |
+
You can usage the model like this:
|
27 |
+
|
28 |
+
```
|
29 |
+
import torch
|
30 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
31 |
+
|
32 |
+
tokenizer = AutoTokenizer.from_pretrained("Prompsit/paraphrase-bert-en")
|
33 |
+
model = AutoModelForSequenceClassification.from_pretrained("Prompsit/paraphrase-bert-en")
|
34 |
+
|
35 |
+
input = tokenizer('may be addressed','could be included',return_tensors='pt')
|
36 |
+
logits = model(**input).logits
|
37 |
+
soft = torch.nn.Softmax(dim=1)
|
38 |
+
print(soft(logits))
|
39 |
+
```
|
40 |
+
Output of previous code is:
|
41 |
+
```
|
42 |
+
tensor([[0.1592, 0.8408]], grad_fn=<SoftmaxBackward>)
|
43 |
+
```
|
44 |
+
As the probability of 1 is 0.84, we can conclude from the previous example that "could be included" is paraphrase of "may be included".
|