Commit
•
da45fba
1
Parent(s):
add6575
Added Infinity endpoint
Browse filesConvert the proposed initial structure to the one compatible with Infinity Endpoint.
- endpoint.py +31 -0
endpoint.py
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import List, Optional, Union
|
2 |
+
|
3 |
+
from infinity.tasks import TextClassificationEndpoint, TextClassificationOutput, \
|
4 |
+
TextClassificationParams
|
5 |
+
from optimum.onnxruntime import ORTModelForSequenceClassification
|
6 |
+
from transformers import pipeline, AutoTokenizer
|
7 |
+
|
8 |
+
class BankingEndpoint(TextClassificationEndpoint):
|
9 |
+
|
10 |
+
__slots__ = ("_pipeline", )
|
11 |
+
|
12 |
+
def __init__(self):
|
13 |
+
super().__init__()
|
14 |
+
|
15 |
+
self._pipeline: Optional[ORTModelForSequenceClassification] = None
|
16 |
+
|
17 |
+
def initialize(self, **kwargs):
|
18 |
+
print("Initializing")
|
19 |
+
model = ORTModelForSequenceClassification.from_pretrained("philschmid/distilbert-onnx-banking77")
|
20 |
+
tokenizer = AutoTokenizer.from_pretrained("philschmid/distilbert-onnx-banking77")
|
21 |
+
|
22 |
+
self._pipeline = pipeline("text-classification", model=model, tokenizer=tokenizer)
|
23 |
+
print("INITIALIZED")
|
24 |
+
|
25 |
+
|
26 |
+
def handle(
|
27 |
+
self,
|
28 |
+
inputs: Union[str, List[str]],
|
29 |
+
parameters: TextClassificationParams
|
30 |
+
) -> List[TextClassificationOutput]:
|
31 |
+
return self._pipeline(inputs, **parameters)
|