Spaces:
Runtime error
Runtime error
Update modeles.py
Browse files- modeles.py +7 -29
modeles.py
CHANGED
@@ -1,41 +1,19 @@
|
|
1 |
-
from transformers import
|
2 |
-
import torch
|
3 |
-
|
4 |
-
def load_and_answer(question, context, model_name):
|
5 |
-
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
6 |
-
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
|
7 |
-
|
8 |
-
# Tokenize the input question-context pair
|
9 |
-
inputs = tokenizer.encode_plus(question, context, max_length=512, truncation=True, padding=True, return_tensors='pt')
|
10 |
-
|
11 |
-
# Send inputs to the same device as your model
|
12 |
-
inputs = {k: v.to(model.device) for k, v in inputs.items()}
|
13 |
-
|
14 |
-
with torch.no_grad():
|
15 |
-
# Forward pass, get model outputs
|
16 |
-
outputs = model(**inputs)
|
17 |
-
|
18 |
-
# Extract the start and end positions of the answer in the tokens
|
19 |
-
answer_start_scores, answer_end_scores = outputs.start_logits, outputs.end_logits
|
20 |
-
answer_start_index = torch.argmax(answer_start_scores) # Most likely start of answer
|
21 |
-
answer_end_index = torch.argmax(answer_end_scores) + 1 # Most likely end of answer; +1 for inclusive slicing
|
22 |
-
|
23 |
-
# Convert token indices to the actual answer text
|
24 |
-
answer_tokens = inputs['input_ids'][0, answer_start_index:answer_end_index]
|
25 |
-
answer = tokenizer.decode(answer_tokens, skip_special_tokens=True)
|
26 |
-
return {"answer": answer, "start": answer_start_index.item(), "end": answer_end_index.item()}
|
27 |
|
28 |
def squeezebert(context, question):
|
29 |
# Define the specific model and tokenizer for SqueezeBERT
|
30 |
model_name = "ALOQAS/squeezebert-uncased-finetuned-squad-v2"
|
31 |
-
|
|
|
32 |
|
33 |
def bert(context, question):
|
34 |
# Define the specific model and tokenizer for BERT
|
35 |
model_name = "ALOQAS/bert-large-uncased-finetuned-squad-v2"
|
36 |
-
|
|
|
37 |
|
38 |
def deberta(context, question):
|
39 |
# Define the specific model and tokenizer for DeBERTa
|
40 |
model_name = "ALOQAS/deberta-large-finetuned-squad-v2"
|
41 |
-
|
|
|
|
1 |
+
from transformers import pipeline
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
def squeezebert(context, question):
|
4 |
# Define the specific model and tokenizer for SqueezeBERT
|
5 |
model_name = "ALOQAS/squeezebert-uncased-finetuned-squad-v2"
|
6 |
+
pip = pipeline('question-answering', model=model_name, tokenizer=model_name)
|
7 |
+
return pip(context=context, question=question)
|
8 |
|
9 |
def bert(context, question):
|
10 |
# Define the specific model and tokenizer for BERT
|
11 |
model_name = "ALOQAS/bert-large-uncased-finetuned-squad-v2"
|
12 |
+
pip = pipeline('question-answering', model=model_name, tokenizer=model_name)
|
13 |
+
return pip(context=context, question=question)
|
14 |
|
15 |
def deberta(context, question):
|
16 |
# Define the specific model and tokenizer for DeBERTa
|
17 |
model_name = "ALOQAS/deberta-large-finetuned-squad-v2"
|
18 |
+
pip = pipeline('question-answering', model=model_name, tokenizer=model_name)
|
19 |
+
return pip(context=context, question=question)
|