The problem about Loading a Pytorch Model in C++

deployment
1

I follow the tutorial and use the same code but it comes up a problem

*** Error in `./example-app': free(): invalid pointer: 0x00007ffe733866b0 ***

I run them in the centos, How can I solve this problem?
2

Could you post the complete stack trace, if you have it?
You are apparently trying to free something which isn’t pointing to a “freeable” memory address.

3

stack trace:

stack%20trace
stack%20trace1694×390 20.7 KB

4

But I only run the same code by following the tutorial.

5

I update the g++ from 4.8.5 to 4.9.2 and it works.

6

I update the g++ too, but still have the same problem.

7

What’s your g++ version?

8

gcc version 4.9.2 20150212 (Red Hat 4.9.2-6) (GCC)

9

You can go to this to explore possible solutions.

10

Sorry, I re-compile the project and no more free() problem.
It is another one :slight_smile:

operation failed in interpreter:
op_version_set = 0
def forward(self,
input: Tensor) -> Tensor:
if bool(torch.gt(torch.sum(input), 0)):
output = torch.mv(self.weight, input)
~~~~~~~~ <— HERE
else:
output_2 = torch.add(self.weight, input, alpha=1)
output = output_2
return output

Abandon (core dumped)

11

Could you please show me your C++ code?

12

I use the tutorial code :

#include <torch/script.h> // One-stop header.

#include
#include

int main(int argc, const char* argv[]) {
if (argc != 2) {
std::cerr << “usage: example-app \n”;
return -1;
}

// Deserialize the ScriptModule from a file using torch::jit::load().
std::shared_ptrtorch::jit::script::Module module = torch::jit::load(argv[1]);

assert(module != nullptr);
std::cout << “ok\n”;
/ Create a vector of inputs.
std::vectortorch::jit::IValue inputs;
inputs.push_back(torch::ones({1, 3, 224, 224}));

// Execute the model and turn its output into a tensor.
at::Tensor output = module->forward(inputs).toTensor();

std::cout << output.slice(/dim=/1, /start=/0, /end=/5) << ‘\n’;
}

13

And I use this code to create the “model.pt”

import torch

import torchvision

An instance of your model.

model = torchvision.models.resnet18()

An example input you would normally provide to your model’s forward() method.

example = torch.rand(1, 3, 224, 224)

print(example)

Use torch.jit.trace to generate a torch.jit.ScriptModule via tracing.

traced_script_module = torch.jit.trace(model, example)

traced_script_module.save(“model.pt”)

14

It has the similar problem with you, you can see this for solution.

15

Thank you. I have the version 1.1.0 but still have the same problem with “torch.mv” function :

terminate called after throwing an instance of ‘std::runtime_error’
what():
mv: Expected 1-D argument vec, but got 4-D (check_1d at /pytorch/aten/src/ATen/native/LinearAlgebra.cpp:143)
frame #0: std::function<std::string ()>::operator()() const + 0x11 (0x7f30c5a46441 in /home/pixur/Documents/Zaynab/TestLibtorch/libtorch/lib/libc10.so)

I didn’t mke any chnge in the code, I just follow the tutorial

16

I used the wrong model.pt. It working now. Thank you!