Hi, I am using the built CPU only docker image as development environment and it seems the module still depends on nvml-go module

sudo docker run --rm -it --name go-onnx-cpu -v $PWD:/go/src/gitlab.com/aiteam c3sr/go-onnxruntime:amd64-cpu-onnxruntime1.7.1-latest

under the environment, when I try to run the go module, this issue shows up. However this is a cpu only onnxruntime build, so we shouldn't be any GPU dependencies here?

/go/pkg/mod/github.com/c3sr/[email protected]/nvml.go:21:11: fatal error: nvml.h: No such file or directory
 // #include <nvml.h>
           ^~~~~~~~
compilation terminated.

My go test code is as follows:

package main

import (
	"context"
	"path/filepath"
	"github.com/c3sr/dlframework/framework/options"
)

var (
	graph_file = "model.onnx"
	batchSize  = 3
)

func main() {
	dir, _ := filepath.Abs("./_fixtures")
	graph := filepath.Join(dir, graph_file)

	device := options.CPU_DEVICE
	ctx := context.Background()

	opts := options.New(options.Context(ctx),
		options.Graph([]byte(graph)),
		options.Device(device, 0),
		options.BatchSize(batchSize))
}

Can't get it to compile on Linux without GPU from official binaries

What I've done

• Downloaded latest .tgz file from onnxruntime releases (onnxruntime-linux-x64-1.9.0.tgz)
• Created empty folder /opt/onnxruntime
• Extracted tgz file into that folder
• I've set up bashrc as described in the instruction

Error

go get -v -tags='nogpu' github.com/c3sr/go-onnxruntime
github.com/c3sr/go-onnxruntime
# github.com/c3sr/go-onnxruntime
In file included from ../go/pkg/mod/github.com/c3sr/[email protected]/device.go:3:
./cbits/predictor.hpp:10:10: fatal error: onnxruntime_c_api.h: No such file or directory
   10 | #include <onnxruntime_c_api.h>
      |          ^~~~~~~~~~~~~~~~~~~~~
compilation terminated.

Bash rc setup

export LIBRARY_PATH=$LIBRARY_PATH:/opt/onnxruntime/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/onnxruntime/lib

Values of variables

echo $LIBRARY_PATH $LD_LIBRARY_PATH
:/opt/onnxruntime/lib :/opt/onnxruntime/lib

How /opt/onnxruntime/ looks

tree /opt/onnxruntime/
/opt/onnxruntime/
├── GIT_COMMIT_ID
├── include
│   ├── cpu_provider_factory.h
│   ├── onnxruntime_c_api.h
│   ├── onnxruntime_cxx_api.h
│   ├── onnxruntime_cxx_inline.h
│   ├── onnxruntime_run_options_config_keys.h
│   ├── onnxruntime_session_options_config_keys.h
│   └── provider_options.h
├── lib
│   ├── libonnxruntime.so -> libonnxruntime.so.1.9.0
│   └── libonnxruntime.so.1.9.0
├── LICENSE
├── Privacy.md
├── README.md
├── ThirdPartyNotices.txt
└── VERSION_NUMBER

I have both the CUDA, cuda driver and cuDNN installed in my ubuntu system. Followed the instructions in c3sr/go-onnxruntime repo for installation

• Successfully installed ONNXRUntime C++ API(built from src with cuda-- built and installed successfully)

• nvidia-smi and nvcc --version works

• can compile the acuda file and run from NVIDIA test example

• g++ version and gcc version are same.

• g++ (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0

• gcc (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0
  Example code that i run is this
  I am continuously getting this error below.
  Undefined variable"NVPA_RawMetricsConfigOptions" in Metric.h
  Am i missing something?
  ` $ go run main.go

  (github.com/c3sr/go-cupti)

In file included from utils.cpp:17:
../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Metric.h: In function ‘bool NV::Metric::Config::GetConfigImage(std::string, std::vector<std::__cxx11::basic_string >, std::vector&, const uint8_t*)’:
../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Metric.h:68:17: error: ‘NVPA_RawMetricsConfigOptions’ was not declared in this scope; did you mean ‘NVPA_RawMetricsConfig’?
68 | NVPA_RawMetricsConfigOptions metricsConfigOptions = { NVPA_RAW_METRICS_CONFIG_OPTIONS_STRUCT_SIZE };
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
| NVPA_RawMetricsConfig
../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Metric.h:69:17: error: ‘metricsConfigOptions’ was not declared in this scope
69 | metricsConfigOptions.activityKind = NVPA_ACTIVITY_KIND_PROFILER;
| ^~~~~~~~~~~~~~~~~~~~
In file included from ../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Metric.h:4,
from utils.cpp:17:
../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Metric.h:72:45: error: ‘NVPA_RawMetricsConfig_Create’ was not declared in this scope; did you mean ‘NVPW_CUDA_RawMetricsConfig_Create’?
72 | RETURN_IF_NVPW_ERROR(false, NVPA_RawMetricsConfig_Create(&metricsConfigOptions, &pRawMetricsConfig));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
../../../go/pkg/mod/github.com/c3sr/[email protected]/csrc/Utils.h:7:26: note: in definition of macro ‘RETURN_IF_NVPW_ERROR’
7 | NVPA_Status status = actual; `

github.com/c3sr/go-onnxruntime

In file included from error.cpp:1:
../../pkg/mod/github.com/c3sr/[email protected]/cbits/error.hpp:4:10: fatal error: 'exceptions.h' file not found

I'm trying to do inference with model generated with Keras and saved to onnx format. But I'm getting an error while creating the model. Code was based on provided example

Error

go run -tags='nogpu' main.go 
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0xcd36ff]

goroutine 1 [running]:
main.main()
        /home/idk/golang_onnx/main.go:27 +0x1ff
exit status 2
(base) idk@idk:~

Code

package main

import (
	"context"
	"fmt"
	"path/filepath"

	"github.com/c3sr/dlframework/framework/options"
	"github.com/c3sr/go-onnxruntime"
	"github.com/c3sr/tracer"
)

func main() {
	device := options.CPU_DEVICE
	graph := filepath.Join("/home/idk/golang_onnx", "model.onnx")

	ctx := context.Background()
	opts := options.New(options.Context(ctx),
		options.Graph([]byte(graph)),
		options.Device(device, 0),
		options.BatchSize(9))

	opts.SetTraceLevel(tracer.FULL_TRACE)

	span, ctx := tracer.StartSpanFromContext(ctx, tracer.FULL_TRACE, "onnxruntime_batch")
	defer span.Finish()

	predictor, err := onnxruntime.New(
		ctx,
		options.WithOptions(opts),
	)
	defer predictor.Close()
	fmt.Println(err)
}