Moaz Reyad created SINGA-404:
--------------------------------
Summary: Call Rafiki from SQL user-defined functions
Key: SINGA-404
URL: https://issues.apache.org/jira/browse/SINGA-404
Project: Singa
Issue Type: Improvement
Reporter: Moaz Reyad
This ticket is to propose a new implementation of the case study defined in the
[Rafiki paper|https://arxiv.org/pdf/1804.06087.pdf]. The case study presented a
user-defined function (UDF) which is called from SQL to send a request to
Rafiki and return the prediction. The SQL user does not need to know how Rafiki
or deep learning works.
Here is a Python implementation of the case study for prediction that works
with the new Rafiki API. The goal is to support the prediction function first,
then later we can add the training function also.
The user defined function in Python:
{code:python}
import json
def predict(query):
url = 'http://127.0.0.1:30000/predict'
headers = {'content-type': 'application/json', 'Accept-Charset':
'UTF-8'}
r = requests.post(url, data=query, headers=headers)
return json.loads(r.text)['prediction']
{code}
Next we add the UDF to SQLite (in memory) database. Let's call it ImageClass
because it returns the class of the image.
{code:python}
import sqlite3
con = sqlite3.connect(":memory:")
con.create_function("ImageClass", 1, predict)
{code}
An example of calling the UDF. Let's create a table and insert a row of data.
This data is taken from the [Rafiki
documentation.|https://nginyc.github.io/rafiki/docs/docs/user/quickstart.html]
{code:python}
cur = con.cursor()
cur.execute("create table test (image varchar(255), author varchar(255))")
cur.execute("insert into test values ('{\"query\": [[0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 1,
0, 0, 7, 0, 37, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 27, 84,
11, 0, 0, 0, 0, 0, 0, 119, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,
0, 88, 143, 110, 0, 0, 0, 0, 22, 93, 106, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 4, 0, 53, 129, 120, 147, 175, 157, 166, 135, 154, 168, 140, 0, 0], [0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 11, 137, 130, 128, 160, 176, 159, 167,
178, 149, 151, 144, 0, 0], [0, 0, 0, 0, 0, 0, 1, 0, 2, 1, 0, 3, 0, 0, 115, 114,
106, 137, 168, 153, 156, 165, 167, 143, 157, 158, 11, 0], [0, 0, 0, 0, 1, 0, 0,
0, 0, 0, 3, 0, 0, 89, 139, 90, 94, 153, 149, 131, 151, 169, 172, 143, 159, 169,
48, 0], [0, 0, 0, 0, 0, 0, 2, 4, 1, 0, 0, 0, 98, 136, 110, 109, 110, 162, 135,
144, 149, 159, 167, 144, 158, 169, 119, 0], [0, 0, 2, 2, 1, 2, 0, 0, 0, 0, 26,
108, 117, 99, 111, 117, 136, 156, 134, 154, 154, 156, 160, 141, 147, 156, 178,
0], [3, 0, 0, 0, 0, 0, 0, 21, 53, 92, 117, 111, 103, 115, 129, 134, 143, 154,
165, 170, 154, 151, 154, 143, 138, 150, 165, 43], [0, 0, 23, 54, 65, 76, 85,
118, 128, 123, 111, 113, 118, 127, 125, 139, 133, 136, 160, 140, 155, 161, 144,
155, 172, 161, 189, 62], [0, 68, 94, 90, 111, 114, 111, 114, 115, 127, 135,
136, 143, 126, 127, 151, 154, 143, 148, 125, 162, 162, 144, 138, 153, 162, 196,
58], [70, 169, 129, 104, 98, 100, 94, 97, 98, 102, 108, 106, 119, 120, 129,
149, 156, 167, 190, 190, 196, 198, 198, 187, 197, 189, 184, 36], [16, 126, 171,
188, 188, 184, 171, 153, 135, 120, 126, 127, 146, 185, 195, 209, 208, 255, 209,
177, 245, 252, 251, 251, 247, 220, 206, 49], [0, 0, 0, 12, 67, 106, 164, 185,
199, 210, 211, 210, 208, 190, 150, 82, 8, 0, 0, 0, 178, 208, 188, 175, 162,
158, 151, 11], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]}', 'Someone')")
{code}
If Rafiki prediction service is running on the specified address, we can call
the Select statement and use Rafiki prediction to return the image class of the
given image.
{code:python}
cur.execute("select author, ImageClass(image) from test")
print(cur.fetchone())
{code}
The result is:
{code:bash}
('Someone', 9)
{code}
More complex Select statements can be written with Joins, Group BY and Order
By, ... etc.
If this implementation is useful, we may add it to Rafiki client as an
experimental feature.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
