BlackToothGrin96
commented
5 years ago Ok, This hasn't been a problem for me; I have been successful in converting the classification.py script into a regression script.
One final question I have is how do I access inference for different models hosted on the same serving instance?
Currently I can access inference via the api endpoints but only for the latest version of the model specified by the ENV MODEL_NAME variable in tfserving dockerfile. I can see through the racket dashboard and racket ls that I have other versions of this model and some completely different models on the server and I'm wondering if I can specify in my request WHICH network should complete the request?
For example, I would like to send inference data to the server with a header specifying which model and version should complete the request.
Description
Hi, I'm loving Racket; it's great work and you are definitely a hero.
I have created two separate projects using the racket init --name command, everything seems to work fine but the file regression.py was not generated for either project (which is disappointing as regression is my goal :( )
What I Did
First Project: $ racket init --name racket_project_0 Folder Tree: . ├── classification.py ├── docker-compose.yaml ├── Dockerfile.racket ├── Dockerfile.tfserving ├── extras │ └── TEST.jpg ├── racket.yaml ├── serialized │ ├── base │ │ └── 1 │ │ ├── saved_model.pb │ │ └── variables │ │ ├── variables.data-00000-of-00001 │ │ └── variables.index │ ├── base_1.h5 │ ├── base_1.json │ └── base_history_1.json ├── web │ ├── asset-manifest.json │ ├── index.html │ ├── manifest.json │ ├── precache-manifest.3be394996438461aed1dfe1534b1c647.js │ ├── service-worker.js │ └── static │ ├── css │ │ ├── 2.280c789f.chunk.css │ │ ├── 2.280c789f.chunk.css.map │ │ ├── main.4cc9d3bc.chunk.css │ │ └── main.4cc9d3bc.chunk.css.map │ └── js │ ├── 2.2580a9e0.chunk.js │ ├── 2.2580a9e0.chunk.js.map │ ├── main.816293ed.chunk.js │ ├── main.816293ed.chunk.js.map │ ├── runtime~main.a8a9905a.js │ └── runtime~main.a8a9905a.js.map └── sqlite.db
Second project: $ racket init --name racket_house_prices_regression Folder Structure: . ├── classification.py ├── docker-compose.yaml ├── Dockerfile.racket ├── Dockerfile.tfserving ├── extras │ └── TEST.jpg ├── racket.yaml ├── serialized │ ├── base │ │ └── 1 │ │ ├── saved_model.pb │ │ └── variables │ │ ├── variables.data-00000-of-00001 │ │ └── variables.index │ ├── base_1.h5 │ ├── base_1.json │ └── base_history_1.json ├── sqlite.db ├── web │ ├── asset-manifest.json │ ├── index.html │ ├── manifest.json │ ├── precache-manifest.3be394996438461aed1dfe1534b1c647.js │ ├── service-worker.js │ └── static │ ├── css │ │ ├── 2.280c789f.chunk.css │ │ ├── 2.280c789f.chunk.css.map │ │ ├── main.4cc9d3bc.chunk.css │ │ └── main.4cc9d3bc.chunk.css.map │ └── js │ ├── 2.2580a9e0.chunk.js │ ├── 2.2580a9e0.chunk.js.map │ ├── main.816293ed.chunk.js │ ├── main.816293ed.chunk.js.map │ ├── runtime~main.a8a9905a.js │ └── runtime~main.a8a9905a.js.map └── housing.csv
Is there anything important I'll be missing by working from classification.py or can I just alter the model definition to suit a regression scenario?
Thanks for the great work btw; this package is awesome :)