Commit
•
e817f0f
1
Parent(s):
d15d6ed
Update README.md
Browse files
README.md
CHANGED
@@ -1,12 +1,41 @@
|
|
1 |
-
#Factual Consistency Evaluator/Metric in ACL 2023 paper
|
2 |
|
3 |
*[WeCheck: Strong Factual Consistency Checker via Weakly Supervised Learning
|
4 |
](https://arxiv.org/abs/2212.10057)*
|
5 |
|
6 |
-
|
7 |
-
WeCheck
|
8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
license: openrail
|
11 |
pipeline_tag: text-classification
|
12 |
tags:
|
|
|
1 |
+
# Factual Consistency Evaluator/Metric in ACL 2023 paper
|
2 |
|
3 |
*[WeCheck: Strong Factual Consistency Checker via Weakly Supervised Learning
|
4 |
](https://arxiv.org/abs/2212.10057)*
|
5 |
|
6 |
+
## Model description
|
7 |
+
WeCheck is a factual consistency metric trained from weakly annotated samples.
|
8 |
|
9 |
+
This WeCheck checkpoint can be used to check the following three generation tasks:
|
10 |
+
|
11 |
+
**Text Summarization/Knowlege grounded dialogue Generation/Paraphrase**
|
12 |
+
|
13 |
+
This WeCheck checkpoint is trained based on the following three weak labler:
|
14 |
+
|
15 |
+
*[QAFactEval
|
16 |
+
](https://github.com/salesforce/QAFactEval)* / *[Summarc](https://github.com/tingofurro/summac)* / *[NLI warmup](https://huggingface.co/MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli)*
|
17 |
---
|
18 |
+
|
19 |
+
### How to use the model
|
20 |
+
```python
|
21 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
22 |
+
import torch
|
23 |
+
|
24 |
+
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
|
25 |
+
model_name = "nightdessert
|
26 |
+
/
|
27 |
+
WeCheck "
|
28 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
29 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
30 |
+
premise = "I first thought that I liked the movie, but upon second thought it was actually disappointing."
|
31 |
+
hypothesis = "The movie was not good."
|
32 |
+
input = tokenizer(premise, hypothesis, truncation=True, return_tensors="pt")
|
33 |
+
output = model(input["input_ids"].to(device)) # device = "cuda:0" or "cpu"
|
34 |
+
prediction = torch.softmax(output["logits"][0], -1).tolist()
|
35 |
+
label_names = ["entailment", "neutral", "contradiction"]
|
36 |
+
prediction = {name: round(float(pred) * 100, 1) for pred, name in zip(prediction, label_names)}
|
37 |
+
print(prediction)
|
38 |
+
|
39 |
license: openrail
|
40 |
pipeline_tag: text-classification
|
41 |
tags:
|