Upload handler.py
Browse files- handler.py +37 -0
handler.py
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# handler.py
|
2 |
+
import torch
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
4 |
+
|
5 |
+
# check for GPU
|
6 |
+
device = 0 if torch.cuda.is_available() else -1
|
7 |
+
|
8 |
+
# multi-model list
|
9 |
+
multi_model_list = [
|
10 |
+
{"model_id": "BAAI/bge-base-en-v1.5", "task":"sentence-embeddings"},
|
11 |
+
{"model_id": "BAAI/bge-reranker-base", "task":"sentence-ranking"},
|
12 |
+
]
|
13 |
+
|
14 |
+
class EndpointHandler():
|
15 |
+
def __init__(self, path=""):
|
16 |
+
self.multi_model={}
|
17 |
+
# load all the models onto device
|
18 |
+
for model in multi_model_list:
|
19 |
+
self.multi_model[model["model_id"]] = pipeline(model["task"], model=model["model_id"], device=device)
|
20 |
+
|
21 |
+
def __call__(self, data):
|
22 |
+
# deserialize incomin request
|
23 |
+
inputs = data.pop("inputs", data)
|
24 |
+
parameters = data.pop("parameters", None)
|
25 |
+
model_id = data.pop("model_id", None)
|
26 |
+
|
27 |
+
# check if model_id is in the list of models
|
28 |
+
if model_id is None or model_id not in self.multi_model:
|
29 |
+
raise ValueError(f"model_id: {model_id} is not valid. Available models are: {list(self.multi_model.keys())}")
|
30 |
+
|
31 |
+
# pass inputs with all kwargs in data
|
32 |
+
if parameters is not None:
|
33 |
+
prediction = self.multi_model[model_id](inputs, **parameters)
|
34 |
+
else:
|
35 |
+
prediction = self.multi_model[model_id](inputs)
|
36 |
+
# postprocess the prediction
|
37 |
+
return prediction
|