Basic example of deploying a clipper server, and deploying an update to an existing prediction model.
# clone this repo
$ git clone [email protected]:mitch-seymour/clipper-demo.git
$ cd clipper-demo
# install the dependencies
$ conda create --name clipper --file requirements.txt- Runs a clipper server locally (server.py)
- Deploys a prediction model (server.py)
- Updates the prediction model (update_model.py)
# activate our anaconda environment for this project
# pro tip, if you setup autoenv, the clipper project should be set automatically
# when you cd into this directory
$ source activate clipper
# install clipper directly from github
$ pip install git+https://github.com/ucbrise/clipper.git@develop#subdirectory=clipper_adminOkay, now start the clipper server. This will take several minutes the first time since the necessary Docker images will need to be installed.
$ python server.pyOnce the server is running, you should be able to hit the query API to run a prediction:
$ curl -X POST --header "Content-Type:application/json" -d '{"input": [1.1, 2.2, 3.3]}' 127.0.0.1:1337/hello-world/predict
# output
{"query_id":3,"output":6.6,"default":false}Notice the output above has a value of 6.6. Now, lets deploy an update to the model, and simply multiply the current output
by 2. In a separate window, run the following:
# always do this first unless you have autoenv installed
$ source activate clipper
# deploy an update to the model
$ python update_model.pyFinally, run the curl command, and notice how the results changed (because our model was updated).
$ curl -X POST --header "Content-Type:application/json" -d '{"input": [1.1, 2.2, 3.3]}' 127.0.0.1:1337/hello-world/predict
# output
{"query_id":4,"output":13.2,"default":false}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent