PereLluis13
commited on
Commit
·
ab249d6
1
Parent(s):
f3a761e
Update README.md
Browse files
README.md
CHANGED
@@ -5,3 +5,106 @@ tags:
|
|
5 |
- seq2seq
|
6 |
license: cc-by-nc-sa-4.0
|
7 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
- seq2seq
|
6 |
license: cc-by-nc-sa-4.0
|
7 |
---
|
8 |
+
To use the model with a pipeline:
|
9 |
+
```python3
|
10 |
+
from transformers import pipeline
|
11 |
+
|
12 |
+
def extract_triplets(text):
|
13 |
+
triplets = []
|
14 |
+
relation = ''
|
15 |
+
for token in text.split():
|
16 |
+
if token == "<triplet>":
|
17 |
+
current = 't'
|
18 |
+
if relation != '':
|
19 |
+
triplets.append((subject, relation, object_))
|
20 |
+
relation = ''
|
21 |
+
subject = ''
|
22 |
+
elif token == "<subj>":
|
23 |
+
current = 's'
|
24 |
+
if relation != '':
|
25 |
+
triplets.append((subject, relation, object_))
|
26 |
+
object_ = ''
|
27 |
+
elif token == "<obj>":
|
28 |
+
current = 'o'
|
29 |
+
relation = ''
|
30 |
+
else:
|
31 |
+
if current == 't':
|
32 |
+
subject += ' ' + token
|
33 |
+
elif current == 's':
|
34 |
+
object_ += ' ' + token
|
35 |
+
elif current == 'o':
|
36 |
+
relation += ' ' + token
|
37 |
+
triplets.append((subject, relation, object_))
|
38 |
+
return triplets
|
39 |
+
|
40 |
+
triplet_extractor = pipeline('text2text-generation', model='Babelscape/rebel-large', tokenizer='Babelscape/rebel-large')
|
41 |
+
extracted_text = triplet_extractor("Punta Cana is a resort town in the municipality of Higüey, in La Altagracia Province, the easternmost province of the Dominican Republic.Punta Cana is a resort town in the municipality of Higüey, in La Altagracia Province, the easternmost province of the Dominican Republic.")["generated_text"]
|
42 |
+
|
43 |
+
extracted_triplets = extract_triplets(extracted_text)
|
44 |
+
print(extracted_triplets)
|
45 |
+
```
|
46 |
+
|
47 |
+
Or using the transformers
|
48 |
+
```python3
|
49 |
+
|
50 |
+
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
|
51 |
+
|
52 |
+
def extract_triplets(text):
|
53 |
+
triplets = []
|
54 |
+
relation = ''
|
55 |
+
for token in text.split():
|
56 |
+
if token == "<triplet>":
|
57 |
+
current = 't'
|
58 |
+
if relation != '':
|
59 |
+
triplets.append((subject, relation, object_))
|
60 |
+
relation = ''
|
61 |
+
subject = ''
|
62 |
+
elif token == "<subj>":
|
63 |
+
current = 's'
|
64 |
+
if relation != '':
|
65 |
+
triplets.append((subject, relation, object_))
|
66 |
+
object_ = ''
|
67 |
+
elif token == "<obj>":
|
68 |
+
current = 'o'
|
69 |
+
relation = ''
|
70 |
+
else:
|
71 |
+
if current == 't':
|
72 |
+
subject += ' ' + token
|
73 |
+
elif current == 's':
|
74 |
+
object_ += ' ' + token
|
75 |
+
elif current == 'o':
|
76 |
+
relation += ' ' + token
|
77 |
+
triplets.append((subject, relation, object_))
|
78 |
+
return triplets
|
79 |
+
|
80 |
+
# Load model and tokenizer
|
81 |
+
tokenizer = AutoTokenizer.from_pretrained("model/rebel-large")
|
82 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("model/rebel-large")
|
83 |
+
gen_kwargs = {
|
84 |
+
"max_length": 256,
|
85 |
+
"length_penalty": 0,
|
86 |
+
"num_beams": 3,
|
87 |
+
"num_return_sequences": 3,
|
88 |
+
}
|
89 |
+
|
90 |
+
# Text to extract triplets from
|
91 |
+
text = 'Punta Cana is a resort town in the municipality of Higüey, in La Altagracia Province, the easternmost province of the Dominican Republic.'
|
92 |
+
|
93 |
+
# Tokenizer text
|
94 |
+
model_inputs = tokenizer(text, max_length=256, padding=True, truncation=True, return_tensors = 'pt')
|
95 |
+
|
96 |
+
# Generate
|
97 |
+
generated_tokens = model.generate(
|
98 |
+
model_inputs["input_ids"].to(model.device),
|
99 |
+
attention_mask=model_inputs["attention_mask"].to(model.device),
|
100 |
+
**gen_kwargs,
|
101 |
+
)
|
102 |
+
|
103 |
+
# Extract text
|
104 |
+
decoded_preds = tokenizer.batch_decode(generated_tokens, skip_special_tokens=False)
|
105 |
+
|
106 |
+
# Extract triplets
|
107 |
+
for idx, sentence in enumerate(decoded_preds):
|
108 |
+
print(f'Prediction triplets sentence {idx}')
|
109 |
+
print(extract_triplets(sentence))
|
110 |
+
```
|