Upload 2 files
Browse files- app___init__.py +0 -0
- app_tapex.py +33 -0
app___init__.py
ADDED
File without changes
|
app_tapex.py
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import TapasTokenizer, TapexTokenizer, BartForConditionalGeneration
|
2 |
+
import pandas as pd
|
3 |
+
import datetime
|
4 |
+
import torch
|
5 |
+
|
6 |
+
|
7 |
+
def execute_query(query, csv_file):
|
8 |
+
a = datetime.datetime.now()
|
9 |
+
|
10 |
+
table = pd.read_csv(csv_file.name, delimiter=",")
|
11 |
+
table = table.astype(str)
|
12 |
+
|
13 |
+
model_name = "microsoft/tapex-large-finetuned-wtq"
|
14 |
+
model = BartForConditionalGeneration.from_pretrained(model_name)
|
15 |
+
tokenizer = TapexTokenizer.from_pretrained(model_name)
|
16 |
+
|
17 |
+
queries = [query]
|
18 |
+
|
19 |
+
encoding = tokenizer(table=table, query=queries, padding=True, return_tensors="pt",truncation=True)
|
20 |
+
outputs = model.generate(**encoding)
|
21 |
+
ans = tokenizer.batch_decode(outputs, skip_special_tokens=True)
|
22 |
+
|
23 |
+
|
24 |
+
|
25 |
+
query_result = {
|
26 |
+
"query": query,
|
27 |
+
"answer": ans[0]
|
28 |
+
}
|
29 |
+
|
30 |
+
b = datetime.datetime.now()
|
31 |
+
print(b - a)
|
32 |
+
|
33 |
+
return query_result, table
|