update readme
Browse files
README.md
CHANGED
@@ -1,3 +1,64 @@
|
|
1 |
---
|
2 |
license: cc-by-sa-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-sa-4.0
|
3 |
+
language: ja
|
4 |
+
pipeline_tag: zero-shot-classification
|
5 |
+
tags:
|
6 |
+
- tohoku-nlp/bert-base-japanese-v3
|
7 |
+
datasets:
|
8 |
+
- shunk031/jsnli
|
9 |
+
library_name: sentence-transformers
|
10 |
---
|
11 |
+
|
12 |
+
|
13 |
+
# Cross-Encoder for Natural Language Inference
|
14 |
+
This model was trained using [SentenceTransformers](https://sbert.net) [Cross-Encoder](https://www.sbert.net/examples/applications/cross-encoder/README.html) class.
|
15 |
+
This model is based on [tohoku-nlp/bert-base-japanese-v3](https://huggingface.co/tohoku-nlp/bert-base-japanese-v3).
|
16 |
+
|
17 |
+
## Training Data
|
18 |
+
The model was trained on the [JSNLI](https://nlp.ist.i.kyoto-u.ac.jp/?%E6%97%A5%E6%9C%AC%E8%AA%9ESNLI%28JSNLI%29%E3%83%87%E3%83%BC%E3%82%BF%E3%82%BB%E3%83%83%E3%83%88) datasets.
|
19 |
+
For a given sentence pair, it will output three scores corresponding to the labels: {0:"entailment", 1:"neutral", 2:"contradiction}.
|
20 |
+
|
21 |
+
## Usage
|
22 |
+
|
23 |
+
Pre-trained models can be used like this:
|
24 |
+
```python
|
25 |
+
from sentence_transformers import CrossEncoder
|
26 |
+
model = CrossEncoder('akiFQC/bert-base-japanese-v3_nli-jsnli')
|
27 |
+
scores = model.predict([('男はピザを食べています', '男は何かを食べています'), ('黒いレーシングカーが観衆の前から発車します。', '男は誰もいない道を運転しています。')])
|
28 |
+
|
29 |
+
#Convert scores to labels
|
30 |
+
label_mapping = ['entailment', 'neutral', 'contradiction',]
|
31 |
+
labels = [label_mapping[score_max] for score_max in scores.argmax(axis=1)]
|
32 |
+
```
|
33 |
+
|
34 |
+
## Usage with Transformers AutoModel
|
35 |
+
You can use the model also directly with Transformers library (without SentenceTransformers library):
|
36 |
+
```python
|
37 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
38 |
+
import torch
|
39 |
+
|
40 |
+
model = AutoModelForSequenceClassification.from_pretrained('cross-encoder/nli-deberta-v3-base')
|
41 |
+
tokenizer = AutoTokenizer.from_pretrained('cross-encoder/nli-deberta-v3-base')
|
42 |
+
|
43 |
+
features = tokenizer(['男はピザを食べています', '黒いレーシングカーが観衆の前から発車します。'], ['男は何かを食べています', '男は誰もいない道を運転しています。'], padding=True, truncation=True, return_tensors="pt")
|
44 |
+
|
45 |
+
model.eval()
|
46 |
+
with torch.no_grad():
|
47 |
+
scores = model(**features).logits
|
48 |
+
label_mapping = ['contradiction', 'entailment', 'neutral']
|
49 |
+
labels = [label_mapping[score_max] for score_max in scores.argmax(dim=1)]
|
50 |
+
print(labels)
|
51 |
+
```
|
52 |
+
|
53 |
+
## Zero-Shot Classification
|
54 |
+
This model can also be used for zero-shot-classification:
|
55 |
+
```python
|
56 |
+
from transformers import pipeline
|
57 |
+
|
58 |
+
classifier = pipeline("zero-shot-classification", model='akiFQC/bert-base-japanese-v3_nli-jsnli')
|
59 |
+
|
60 |
+
sent = "Appleは先程、iPhoneの最新機種について発表しました。"
|
61 |
+
candidate_labels = ["技術", "スポーツ", "政治"]
|
62 |
+
res = classifier(sent, candidate_labels)
|
63 |
+
print(res)
|
64 |
+
```
|