Seldon engine
Machine learning deployment for enterprise
Syndicai allows you to deploy models prepared in Seldon convention so that you can easily deploy your models in a serverless way without any infrastructure setup.
In order to correctly deploy a model, you need to name the main file syndicai.py and the main class syndicai to run a model. Explore different input / output types that are supported by Seldon. For more information, please go to Seldon Docs.

Directory structure

With Seldon approach 2 files are required to deploy a model.
1
./seldon-model/
2
├── syndicai.py
3
├── ...
4
└── requirements.txt
Copied!
  • syndicai.py - Main file responsible for the model run which consists of the python class syndicai and method predict. Depends on the structure of the input data X variable can be either ndarray, JSON, and string.
  • requirements.txt - is the file needed to recreate the environment for the model. It is important to give information about the exact version of the library.
syndicai.py
requirements.txt
1
class syndicai(object):
2
3
def __init__(self, config):
4
pass
5
6
def predict(self, X, features_names=None):
7
""" Model Run function """
8
9
# Your code
10
11
return output
Copied!
1
tensorflow==1.15.0
2
pylab-sdk==1.1.2
Copied!

Sample models

  • Object detector - Classical object detection algorithm based on MobileNet V2. For this task Tensorflow library and Seldon Core was used.
  • Yolo V5 - One of the most popular and powerful computer vision algorithms in the field of real-time object detection. Model written in PyTorch framework.
Last modified 11mo ago