Package {nativeORT}

Type: Package
Title: Native 'R' 'ONNX' Runtime
Version: 0.1.2
Description: Provides 'R' native 'ONNX' model inference without requiring 'Python', 'reticulate' bindings, or 'TensorFlow'. This package directly binds the 'ONNX Runtime' C API via 'Rcpp', enabling real-time inference for '.onnx' engines, all within 'R'. Standard CPU execution is supported as well as the 'CoreML' Execution Provider (CEP) for Apple Silicon, all without external bindings. This package handles OS detection, linking 'ONNX' libraries, and inference. For more information about 'ONNX Runtime' see https://onnxruntime.ai/.
License: MIT + file LICENSE
Language: en-US
Encoding: UTF-8
LinkingTo: Rcpp
Imports: Rcpp, digest, glue
Config/roxygen2/version: 8.0.0
SystemRequirements: libonnxruntime (>= 1.20)
Suggests: ggplot2, knitr, rmarkdown
VignetteBuilder: knitr
URL: https://github.com/calebmcarr/nativeORT
BugReports: https://github.com/calebmcarr/nativeORT/issues
NeedsCompilation: yes
Packaged: 2026-05-07 01:44:21 UTC; calebcarr
Author: Caleb Carr [aut, cre]
Maintainer: Caleb Carr <calebmcarr1@gmail.com>
Repository: CRAN
Date/Publication: 2026-05-12 19:20:08 UTC

ort_binary_url

Description

ort_binary_url

Usage

ort_binary_url()

Value

string where to download onnxruntime from

Examples

ort_binary_url()

ort_codesign

Description

for macs, signs the downloaded binaries

Usage

ort_codesign(lib_dir)

Arguments

lib_dir

where the libraries are

Value

invisible lib path

Examples


  if(!ort_is_installed()){
    ort_codesign(file.path(ort_install_dir(), "lib"))
  }

ort_detect_os

Description

ort_detect_os

Usage

ort_detect_os()

Value

str - platform

Examples

ort_detect_os()

ort_download

Description

ort_download

Usage

ort_download(url, dest_dir)

Arguments

url

where to download from

dest_dir

where to download to

Value

invisible dest_dir

Examples


if (!ort_is_installed()){
  ort_download(ort_binary_url(), ort_install_dir())}

ort_infer_raw

Description

ort_infer_raw

Usage

ort_infer_raw(session, input)

Arguments

session

ORT session

input

tensor to infer on

Value

ORT output

Examples


  if (file.exists("model.onnx")) {
    session <- ort_session("model.onnx")
    input <- array(runif(1 * 3 * 256 * 256), dim = c(1L, 3L, 256L, 256L))
    ort_infer_raw(session, input)
  }

ort_install

Description

ort_install

Usage

ort_install()

Value

invisible where it was saved to

Examples


  if(!ort_is_installed()){
    ort_install()}

ort_install_dir

Description

ort_install_dir

Usage

ort_install_dir()

Value

where the install lives

Examples

ort_install_dir()

ort_is_installed

Description

ort_is_installed

Usage

ort_is_installed()

Value

boolean for if you have onnx runtime or not

Examples

ort_is_installed()

session.R

Description

session.R

Usage

ort_session(
  path,
  type = "detection",
  provider = "cpu",
  cache_dir = NULL,
  threads = 0L,
  opt_level = 99L
)

Arguments

path

where the .onnx is located

type

detection/classificaton/segmentation

provider

cpu or coreml (only for apple silicon)

cache_dir

where coreml cache be directed

threads

= 0 for all, integer otherwise

opt_level

= 99 for all ops, 1 for basics

Value

ORT session object

Examples


  if (file.exists('./yolov11n.onnx')) {
    ort_session('./yolov11n.onnx', 'detection')
  }

print.ort_session

Description

print.ort_session

Usage

## S3 method for class 'ort_session'
print(x, ...)

Arguments

x

session object

...

extra params

Value

invisible session

Examples


  if (file.exists("model.onnx")) {
    session <- ort_session("model.onnx")
    print(session)
  }