mirror of
https://github.com/hanzoai/onnxgo.git
synced 2026-10-30 18:26:28 +00:00
zeekay
349ed099f6
The import path a reader types is the one the org actually owns, rather than the forge that happens to host the bytes. hanzo.ai serves the go-import meta and points it at the same repository, so nothing about where the code lives changes. Rewritten by ROLE, not by string: the module line, the import in each code sample, and the pkg.go.dev address the docs now live at. The two `hanzoai/onnxgo-examples` links stay — that is a different repository, and a plain substitution would have invented `hanzo.ai/onnxgo-examples`, which is not a module at all. The `// Deprecated:` notice belongs to v1.31.1 and does not travel here: it is what the OLD path says about the move, and this commit is no longer that path.
679 B
679 B
hanzoai/onnxgo — ONNX Runtime Go Bindings
High-performance Go bindings for Microsoft ONNX Runtime. GPU inference (CUDA, TensorRT, CoreML) for production AI workloads.
Usage
import ort "hanzo.ai/onnxgo"
ort.SetSharedLibraryPath("/usr/local/lib/libonnxruntime.so")
ort.InitializeEnvironment()
defer ort.DestroyEnvironment()
session, _ := ort.NewAdvancedSession("model.onnx",
[]string{"input"}, []string{"output"},
ort.NewSessionOptions(),
)
defer session.Destroy()
GPU Support
NVIDIA CUDA + cuDNN, TensorRT, Apple CoreML, DirectML, OpenVINO, ROCm.
Origin
Forked from yalue/onnxruntime_go (2026-03-29). Rebranded for Hanzo AI infrastructure.