Cilia Madani
commited on
Commit
•
6edbbf1
1
Parent(s):
c7f37ed
first app
Browse files- biogpt_qa_pubmedqa_biogpt_large.py +81 -0
- requirements.txt +4 -0
biogpt_qa_pubmedqa_biogpt_large.py
ADDED
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""BioGPT-QA-PubMedQA-BioGPT-Large.ipynb
|
3 |
+
|
4 |
+
Automatically generated by Colab.
|
5 |
+
|
6 |
+
Original file is located at
|
7 |
+
https://colab.research.google.com/drive/1grmKyUETABgEsC3hO7CKgnSG1JGBbmHV
|
8 |
+
"""
|
9 |
+
|
10 |
+
|
11 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
12 |
+
import gradio as gr
|
13 |
+
import torch
|
14 |
+
import re
|
15 |
+
|
16 |
+
tokenizer = AutoTokenizer.from_pretrained("microsoft/BioGPT-Large-PubMedQA")
|
17 |
+
model = AutoModelForCausalLM.from_pretrained("microsoft/BioGPT-Large-PubMedQA")
|
18 |
+
|
19 |
+
# Check if GPU is available and move model to GPU
|
20 |
+
if torch.cuda.is_available():
|
21 |
+
device = torch.device("cuda")
|
22 |
+
model.to(device)
|
23 |
+
print("Using GPU:", torch.cuda.get_device_name(0))
|
24 |
+
else:
|
25 |
+
device = torch.device("cpu")
|
26 |
+
print("Using CPU")
|
27 |
+
|
28 |
+
def answer_bio_question(question, context):
|
29 |
+
"""Answers a biological question using BioGPT-QA-PubMedQA-BioGPT-Large.
|
30 |
+
|
31 |
+
Args:
|
32 |
+
question: The question to be answered.
|
33 |
+
context: The context or passage containing the answer (concatenated with the question).
|
34 |
+
|
35 |
+
Returns:
|
36 |
+
The answer extracted from the context based on the question.
|
37 |
+
"""
|
38 |
+
try:
|
39 |
+
# Concatenate question and context with a separator
|
40 |
+
input_ids = tokenizer(question + " [SEP] " + context, return_tensors="pt").to(device)
|
41 |
+
|
42 |
+
# Move the input to the chosen device (GPU or CPU)
|
43 |
+
input_ids = input_ids.to(device)
|
44 |
+
|
45 |
+
outputs = model.generate(**input_ids,
|
46 |
+
max_new_tokens=1024
|
47 |
+
,
|
48 |
+
num_beams=1,
|
49 |
+
early_stopping=False,
|
50 |
+
do_sample=False,
|
51 |
+
)
|
52 |
+
|
53 |
+
|
54 |
+
answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
55 |
+
print(answer)
|
56 |
+
segs = re.search(r"the answer to the question given the context is(.*)", answer)
|
57 |
+
ans = "unknown"
|
58 |
+
if segs is not None:
|
59 |
+
segs = segs.groups()
|
60 |
+
ans = segs[0].strip()
|
61 |
+
|
62 |
+
return "The answer to this question is " + ans
|
63 |
+
except Exception as e:
|
64 |
+
print(f"Error during question answering: {e}")
|
65 |
+
return "An error occurred during question answering. Please try again."
|
66 |
+
|
67 |
+
|
68 |
+
|
69 |
+
iface = gr.Interface(
|
70 |
+
fn=answer_bio_question,
|
71 |
+
inputs=[
|
72 |
+
gr.Textbox(label="Question", lines=2),
|
73 |
+
gr.Textbox(label="Context (Passage)", lines=5),
|
74 |
+
],
|
75 |
+
outputs="textbox",
|
76 |
+
title="BioGPT-QA: Answer Biological Questions",
|
77 |
+
)
|
78 |
+
|
79 |
+
# Launch the Gradio interface
|
80 |
+
iface.launch( debug=True, share=True)
|
81 |
+
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
transformers
|
2 |
+
torch
|
3 |
+
gradio
|
4 |
+
sacremoses
|