I am having a package with its own cmake. I am trying to link the libraries of libtorch with it.
But after successful cmake, following error occured
Error: macro "Error" passed 4 arguments, but takes just 1
AT_ASSERT(impl == nullptr || impl->type() == device.type());
^

Hi there,
I tried to build a minimal C++ based on the latest nightly build of libtorch but got the following errors. Can someone help?
/lib/../lib64/crt1.o: In function _start': (.text+0x20): undefined reference to main'
.../libtorch_nightly/lib/libtorch.so: undefined reference to lgammaf@GLIBC_2.23' .../libtorch_nightly/lib/libtorch.so: undefined reference to lgamma@GLIBC_2.23'
collect2: error: ld returned 1 exit status
gmake[3]: *** [stylegan] Error 1
gmake[2]: *** [CMakeFiles/mymodel.dir/all] Error 2
gmake[1]: *** [CMakeFiles/mymodel.dir/rule] Error 2
gmake: *** [mymodel] Error 2

I was reading the docs, and I feel like here are some things that can be added, and since I'm working, I reaaaallly need it!

Here goes the list:

• Ability to configure src/ as the source folder for source files
• Ability to config include/ for user created Header files
• Docs to install and unzip the Pytorch, instead of manually wgetting it, and unzipping

If these can be added, or I can be guide, that would be awesome!!

Regards,
Have a great day!

hey, thanks for this great framework and the c++ API design philosophy is really user-friendly for old pytorch user using python.

could you please provide some more concrete examples about
(1) how to install and use libtorch not only with cmake ( add g++ example?);
(2) what is the best practice to use pybind to invoke c++ tensor functions or models in python.
considering some pythoners do have the desire to use libtorch and conduct mix programming but with knowing a little about the c++.

@goldsborough @soumith @apaszke

have a nice day.

Hi,

In index.html (https://pytorch.org/cppdocs/frontend.html#installation), I couldn't jump to installation page and found that link is https://pytorch.org/cppdocs/installation.html not https://pytorch.org/cppdocs/installing.html.

As discussed in https://discuss.pytorch.org/t/can-not-access-torch-c-api-documentation/156052

The c++ doc pages can not be opened.

Hey, I just did a simple integration of libtorch as an addon for nodejs. The API worked very nice and I really enjoyed the design, it's very well done ! I'm adding in this issue some feedbacks (don't know if here is the best place, but here you go):

• The torch::jit::load isn't exposed by just adding the torch/torch.h header, you have to include the torch/script.h as well, don't know if it is intentional but it took me some time to find the right header, maybe a documentation improvement could help;
• There are some local paths hardcoded into the ATen cmake files, such as in share/cmake/ATen/ATenConfig.cmake: /Users/administrator/nightlies/2018_10_01/wheel_build_dirs/libtorch_2.7/pytorch/torch/lib/tmp_install/include;
• Documentation is still missing some parts, especially regarding loading TorchScript, etc. Given that a lot of people will start to use the libtorch, I would be nice to have a expanded documentation on ATen/Torch API with some examples (I can contribute as well);
• Problem when seeing the docs for the torch::jit::load (https://pytorch.org/cppdocs/api/function_namespacetorch_1_1jit_1ace2c44fb8af5905ae17834e81086b8a3.html#exhale-function-namespacetorch-1-1jit-1ace2c44fb8af5905ae17834e81086b8a3)
• Many broken broken in the initial page of the docs (torch::nn, torch::optim, torch::data, torch::serialize, torch::jit and torch::python)

This is what I remember, if I see anything else I'll add it here as well.

Hi, apologies if this is not an issue. I walked through the "Minimal Example" for Installing c++ Distributions of PyTorch, and I get through it with no issues. However, when I create the example-app.cpp file, clang throws an error 'torch/torch.h' file not found [clang: pp_file_not_found] (while in neovim) and this causes other issues in the file since it's not reading the next line #include <iostream>. Do I need to place libtorch somewhere within the project or somewhere else for it to be recognized by clang?

Hi, I want to deploy a detection model with TensoRT to accelerate the backbone and use libtorch to do detection(maily reshape and decode boxes) and nms, but how to compiler them together?
When I try to compiler them, many redefined class error appears. Please help me out. Thanks.
A minimal makefile example is appreciated. Thanks.

I have a Lapop with a nvidia graphicscard and I'm trying to use pytorch for cuda with g++. But i couldn't find any good information about dependecies e.g and my compiler always trohws errors, I'm currently using this command I found on the internet: "g++ -std=c++14 main.cpp -I ${TORCH_DIR}/include/torch/csrc/api/include/ -I ${TORCH_DIR}/include/ -L ${TORCH_DIR}/lib/ -L /usr/local/cuda/lib64 -L /usr/local/cuda/nvvm/lib64 -ltorch -lc10 -lc10_cuda -lnvrtc -lcudart_static -ldl -lrt -pthread -o out", but it just says: "torch/torch.h: file not found"

Hi.
I'm trying to convert my python model to C++.
But while doing so, when i call one class from another class, how do i register the class module.

class A(nn.Module):
def init(self, num_features):
super(A, self).init()
self.bn = BatchNorm2d(num_features)

def forward(self, x):
    bn = self.bn(x)
    return bn

def __call__(self, x):
    return self.forward(x)

class B(nn::Module):
def init(self, num_features)
super(BNCN_Gate, self).init()
self.bn = CustomBatchNorm(in_channels, renorm, arg_ex)
conv_gate = nn::sequential()
......`

How should write such using Pytorch C++ API.
Right now i'm able to do this:

`struct A : torch::nn::Module {
A(int64_t num_features) :
bn(torch::nn::BatchNorm(num_features)) {
}
torch::Tensor forward(torch::Tensor x) {
x = bn->forward(x);
return x;
}
torch::nn::BatchNorm bn;
};

struct BNCC_Gate : torch::nn::Module {
BNCC_Gate()
{
bn(CustomBatchNorm(3, NULL, NULL));
this->register_module("bn", bn);
conv_gate = torch::nn::Sequential();
......
}
};
But this is giving me error as: undefined bn.
How should i create a object of the class A and regsiter it to get the complete network structure in class B?

@goldsborough
@soumith @asmeurer @asuhan @xuhdev
Any help would be great.
Thank you