Commit
•
4865f8a
1
Parent(s):
770e1d9
Update pipeline.py
Browse files- pipeline.py +5 -13
pipeline.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
-
from typing import Dict,
|
2 |
from optimum.onnxruntime import ORTModelForFeatureExtraction
|
3 |
-
from transformers import
|
4 |
|
5 |
def cls_pooling(model_output):
|
6 |
return model_output.last_hidden_state[:,0]
|
@@ -11,20 +11,12 @@ class PreTrainedPipeline():
|
|
11 |
self.tokenizer = AutoTokenizer.from_pretrained(path, model_max_length=128)
|
12 |
|
13 |
|
14 |
-
def __call__(self, inputs: Any) -> Dict[str,
|
15 |
-
"""
|
16 |
-
Args:
|
17 |
-
data (:obj:`str`):
|
18 |
-
a string containing some text
|
19 |
-
Return:
|
20 |
-
A :obj:`list`:. The object returned should be a list of one list like [[{"label": 0.9939950108528137}]] containing :
|
21 |
-
- "label": A string representing what the label/class is. There can be multiple labels.
|
22 |
-
- "score": A score between 0 and 1 describing how confident the model is for this label/class.
|
23 |
-
"""
|
24 |
# tokenize the input
|
25 |
encoded_input = self.tokenizer(inputs, padding="longest", truncation=True, return_tensors='pt')
|
26 |
# run the model
|
27 |
model_output = self.model(**encoded_input, return_dict=True)
|
28 |
embeddings = cls_pooling(model_output)
|
29 |
|
30 |
-
return {"vectors":
|
|
|
|
1 |
+
from typing import Dict,List, Any
|
2 |
from optimum.onnxruntime import ORTModelForFeatureExtraction
|
3 |
+
from transformers import AutoTokenizer
|
4 |
|
5 |
def cls_pooling(model_output):
|
6 |
return model_output.last_hidden_state[:,0]
|
|
|
11 |
self.tokenizer = AutoTokenizer.from_pretrained(path, model_max_length=128)
|
12 |
|
13 |
|
14 |
+
def __call__(self, inputs: Any) -> Dict[str, List[float]]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
# tokenize the input
|
16 |
encoded_input = self.tokenizer(inputs, padding="longest", truncation=True, return_tensors='pt')
|
17 |
# run the model
|
18 |
model_output = self.model(**encoded_input, return_dict=True)
|
19 |
embeddings = cls_pooling(model_output)
|
20 |
|
21 |
+
return {"vectors": embeddings[0].tolist()}
|
22 |
+
|