Skip to content

nayotta/fission-python-env-with-pytorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Fission: Python Environment

This is the Python environment for Fission.

It's a Docker image containing a Python 3.5 runtime, along with a dynamic loader. A few common dependencies are included in the requirements.txt file.

Looking for ready-to-run examples? See the Python examples directory.

Customizing this image

To add package dependencies, edit requirements.txt to add what you need, and rebuild this image (instructions below).

You also may want to customize what's available to the function in its request context. You can do this by editing server.py (see the comment in that file about customizing request context).

Rebuilding and pushing the image

You'll need access to a Docker registry to push the image: you can sign up for Docker hub at hub.docker.com, or use registries from gcr.io, quay.io, etc. Let's assume you're using a docker hub account called USER. Build and push the image to the the registry:

   docker build -t USER/python-env . && docker push USER/python-env

Using the image in fission

You can add this customized image to fission with "fission env create":

   fission env create --name python --image USER/python-env

Or, if you already have an environment, you can update its image:

   fission env update --name python --image USER/python-env   

After this, fission functions that have the env parameter set to the same environment name as this command will use this environment.

Web Server Framework

Python environment build and start a WSGI server, to support high HTTP traffic. As it is applied in different use cases, this provides two server frameworks: bjoern and gevent. They all support high concurrency request.

bjoern has good performance on RPS, to be ideal for most light resource utilization cases.

gevent is a good supplement because of its internal multi-threads. It supports heavy resource load functions, with well distribution of response time.

Python environment pod remains bjoern framework by default. And it runs gevent framework by setting the container environment WSGI_FRAMEWORK value to GEVENT.

The environment value is configured normally in two ways. One way is to set in Dockerfile and build it into image. The other way is to set in Kubernetes deployment spec during pod running and restart it.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published