osliusarenko
commited on
Commit
•
d65cf4e
1
Parent(s):
a212615
Update README.md
Browse files
README.md
CHANGED
@@ -17,24 +17,19 @@ This is a baseline RoBERTa-base model for the delicate text detection task.
|
|
17 |
Here's a short usage example with the torch library in a binary classification task:
|
18 |
|
19 |
```python
|
20 |
-
from transformers import
|
21 |
-
import torch
|
22 |
|
23 |
-
|
24 |
-
model = AutoModelForSequenceClassification.from_pretrained("grammarly/detexd-roberta")
|
25 |
-
model.eval()
|
26 |
|
27 |
-
def predict_binary_score(text: str
|
28 |
-
|
29 |
-
|
30 |
-
logits = model(**tokenizer(text, return_tensors='pt'))[0]
|
31 |
-
probs = torch.nn.functional.softmax(logits, dim=-1)
|
32 |
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
|
39 |
def predict_delicate(text: str, threshold=0.72496545):
|
40 |
return predict_binary_score(text) > threshold
|
|
|
17 |
Here's a short usage example with the torch library in a binary classification task:
|
18 |
|
19 |
```python
|
20 |
+
from transformers import pipeline
|
|
|
21 |
|
22 |
+
classifier = pipeline("text-classification", model="grammarly/detexd-roberta-base")
|
|
|
|
|
23 |
|
24 |
+
def predict_binary_score(text: str):
|
25 |
+
# get multiclass probability scores
|
26 |
+
scores = classifier(text, top_k=None)
|
|
|
|
|
27 |
|
28 |
+
# convert to a single score by summing the probability scores
|
29 |
+
# for the higher-index classes
|
30 |
+
return sum(score['score']
|
31 |
+
for score in scores
|
32 |
+
if score['label'] in ('LABEL_3', 'LABEL_4', 'LABEL_5'))
|
33 |
|
34 |
def predict_delicate(text: str, threshold=0.72496545):
|
35 |
return predict_binary_score(text) > threshold
|