Spaces:
Paused
Paused
CountingMstar
commited on
Commit
•
32a4ae2
1
Parent(s):
89057d0
Update app.py
Browse files
app.py
CHANGED
@@ -10,20 +10,20 @@ tokenizer = BertTokenizerFast.from_pretrained('bert-base-uncased')
|
|
10 |
model = BertForQuestionAnswering.from_pretrained("CountingMstar/ai-tutor-bert-model")
|
11 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
12 |
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
|
17 |
-
|
18 |
-
|
19 |
|
20 |
-
|
21 |
|
22 |
-
|
23 |
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
|
28 |
def split(text):
|
29 |
context, question = '', ''
|
@@ -44,14 +44,14 @@ def split(text):
|
|
44 |
|
45 |
return context[:-2], question[1:]
|
46 |
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
def greet(text):
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
|
56 |
iface = gr.Interface(fn=greet, inputs="text", outputs="text")
|
57 |
iface.launch()
|
|
|
10 |
model = BertForQuestionAnswering.from_pretrained("CountingMstar/ai-tutor-bert-model")
|
11 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
12 |
|
13 |
+
def get_prediction(context, question):
|
14 |
+
inputs = tokenizer.encode_plus(question, context, return_tensors='pt').to(device)
|
15 |
+
outputs = model(**inputs)
|
16 |
|
17 |
+
answer_start = torch.argmax(outputs[0])
|
18 |
+
answer_end = torch.argmax(outputs[1]) + 1
|
19 |
|
20 |
+
answer = tokenizer.convert_tokens_to_string(tokenizer.convert_ids_to_tokens(inputs['input_ids'][0][answer_start:answer_end]))
|
21 |
|
22 |
+
return answer
|
23 |
|
24 |
+
def question_answer(context, question):
|
25 |
+
prediction = get_prediction(context,question)
|
26 |
+
return prediction
|
27 |
|
28 |
def split(text):
|
29 |
context, question = '', ''
|
|
|
44 |
|
45 |
return context[:-2], question[1:]
|
46 |
|
47 |
+
def greet(texts):
|
48 |
+
context, question = split(texts)
|
49 |
+
answer = question_answer(context, question)
|
50 |
+
return answer
|
51 |
+
# def greet(text):
|
52 |
+
# context, question = split(text)
|
53 |
+
# # answer = question_answer(context, question)
|
54 |
+
# return context
|
55 |
|
56 |
iface = gr.Interface(fn=greet, inputs="text", outputs="text")
|
57 |
iface.launch()
|