In botchkoAI/vertaReticulateClient: Reticulated-based client to the Verta AI Platform

knitr::opts_chunk$set(echo = TRUE)

Poisson Regression Example

In this vignette we'll build a simple model from rglm's help function and deploy it to Verta.

Building the modell

## Dobson (1990) Page 93: Randomized Controlled Trial :
counts <- c(18,17,15,20,10,20,25,13,12)
outcome <- gl(3,1,9)
treatment <- gl(3,3)
original_df <- data.frame(treatment, outcome, counts) # showing data
original_df

glm.D93 <- glm(counts ~ outcome + treatment, family = poisson())

Test predictions with a simple datapoint.

predict(glm.D93,newdata = data.frame(outcome ="1",treatment  = "2" ))

Prepare the Prediciton object for Deployment

In this case we simply pass the object. The other parameters are not actually needed. The structure of the object is as follows: * model_file - the object we call the rpredict function on. * formula_obj - optional,

poisson_model_object <- list(
  model_file = glm.D93
)

saveRDS(poisson_model_object,"poisson_model_object.RData")

Set up Verta Connection

library(stringr)
reg_random_string <- rpois(1,1000)
experiment_random_string <- rpois(1,1000)

reg_random_string <- rpois(1,1000)
experiment_random_string <- rpois(1,1000)

PROJECT_NAME <-  str_glue("Poisson GLM Example {reg_random_string}")
EXPERIMENT_NAME <-  str_glue("Poisson GLM Example {experiment_random_string}")
WORKSPACE <-  "Stefan_Petrov"

# library(vertaReticulateClient)
HOST = "dev.verta.ai"
Sys.setenv(VERTA_EMAIL= "...",
          VERTA_DEV_KEY = "...")

library(vertaReticulateClient)
init_verta(HOST)
proj <- set_verta_project(name=PROJECT_NAME)
experiment <- set_experiment(name=EXPERIMENT_NAME)

run <- set_experiment_run(name=EXPERIMENT_NAME,desc = "Poisson Model Run")
# run$log_model(model = "poisson_model_object.RData",overwrite = T)# # TODO- add# run_log
log_model(run, "poisson_model_object.RData", overwrite = T)

Download the model

run$download_model("./poisson_model_object_downloaded.RData")

downloaded_model <- readRDS("poisson_model_object_downloaded.RData")# this
downloaded_model 

Build the Docker image

createDockerContextZip(
  "poisson_model_object_downloaded.RData",
  required_packages = c() 
)

buildDocker(location = "docker_folder_WZRUV5146S/",tag="verta-poisson:latest")

run it

docker kill $(docker container ls -q)
docker run -p 8000:8000 verta-poisson:latest

Example http call and response:

Post to verta-poisson

POST http://localhost:8000/score Content-Type: application/json

[{"treatment":"1","outcome": "1"}]

Response

HTTP/1.1 200 OK Date: Fri, 30 Jul 2021 17:03:38 GMT Content-Type: application/json Content-Length: 8

[ 3.0445 ]

Response code: 200 (OK); Time: 143ms; Content length: 8 bytes

botchkoAI/vertaReticulateClient documentation built on Dec. 19, 2021, 10:50 a.m.