jonathanjordan21
commited on
Commit
•
b01c113
1
Parent(s):
bccc14d
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,57 @@
|
|
1 |
from fastapi import FastAPI
|
|
|
|
|
2 |
|
3 |
app = FastAPI()
|
4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
@app.get("/")
|
6 |
def greet_json():
|
7 |
return {"Hello": "World!"}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from fastapi import FastAPI
|
2 |
+
from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
|
3 |
+
import torch
|
4 |
|
5 |
app = FastAPI()
|
6 |
|
7 |
+
|
8 |
+
|
9 |
+
model_name = "cardiffnlp/twitter-xlm-roberta-base-sentiment"
|
10 |
+
sentiment_model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
11 |
+
sentiment_tokenizer = AutoTokenizer.from_pretrained(model_name)
|
12 |
+
sentiment_model.config.id2label[3] = "mixed"
|
13 |
+
|
14 |
@app.get("/")
|
15 |
def greet_json():
|
16 |
return {"Hello": "World!"}
|
17 |
+
|
18 |
+
|
19 |
+
@app.post(/sentiment_score)
|
20 |
+
async def sentiment_score(text: str):
|
21 |
+
inputs = sentiment_tokenizer(text[:2500], return_tensors='pt')
|
22 |
+
|
23 |
+
with torch.no_grad():
|
24 |
+
logits = sentiment_model(**inputs).logits #+ 1
|
25 |
+
|
26 |
+
|
27 |
+
print(logits)
|
28 |
+
|
29 |
+
logits = logits + logits[0,1].abs()
|
30 |
+
|
31 |
+
# print(torch.nn.functional.sigmoid(logits))
|
32 |
+
|
33 |
+
# logits = logits / 10
|
34 |
+
|
35 |
+
# print(logits)
|
36 |
+
|
37 |
+
# print(torch.abs(logits[0,0] - logits[0,-1]))
|
38 |
+
# print(logits[0,1]//torch.max(torch.abs(logits[0,::2])))
|
39 |
+
|
40 |
+
logits = torch.cat(
|
41 |
+
(
|
42 |
+
logits, (
|
43 |
+
# ( logits[0,1] + torch.sign(logits[0,0] - logits[0,-1]) * (logits[0,0] - logits[0,-1])/2 )/2 +
|
44 |
+
# (logits[0,0] + logits[0,-1])/20
|
45 |
+
(1 - torch.abs(logits[0,0] - logits[0,-1])*(2+(logits[0,1]//torch.max(torch.abs(logits[0,::2])))))
|
46 |
+
).unsqueeze(0).unsqueeze(0)
|
47 |
+
), dim=-1
|
48 |
+
)
|
49 |
+
|
50 |
+
softmax = torch.nn.functional.softmax(
|
51 |
+
logits,
|
52 |
+
dim=-1
|
53 |
+
)
|
54 |
+
|
55 |
+
return [{"label":model.config.id2label[predicted_class_id.tolist()], "score":softmax[0, predicted_class_id].tolist()} for predicted_class_id in softmax.argsort(dim=-1, descending=True)[0]]
|
56 |
+
|
57 |
+
|