STATUS_DLL_NOT_FOUND when using LIBTORCH_USE_PYTORCH

[Daan v O]

Please, this is a pain to work with.

I am using version tch = "0.13.0"
Just the normal starting code:

use tch::Tensor;

fn main() {
    let t = Tensor::from_slice(&[3, 1, 4, 1, 5]);
    let t = t * 2;
    t.print();
}

Installed PyTorch within python:

Python 3.11.6 (tags/v3.11.6:8b6ee5b, Oct  2 2023, 14:57:12) [MSC v.1935 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.__version__)
2.0.0+cpu

I set my enviroment variable LIBTORCH_USE_PYTORCH to 1.

This are all the steps I take, do I miss a step? I see some people talk about setting environment variables, but not which ones.

I build and run:

cargo run
    Finished dev [unoptimized + debuginfo] target(s) in 0.04s
     Running `target\debug\test-libtorch.exe`
error: process didn't exit successfully: `target\debug\test-libtorch.exe` (exit code: 0xc0000135, STATUS_DLL_NOT_FOUND)

I tried to install the libtorch zip and set the environment variable and path, even though I thought that shouldn't be necessary. And the only version from libtorch I could find is 2.2.0, not the right one. So I removed this as well.

[Christopher Fleetwood]

Maybe useful to use Dependency Walker to determine the exact DLL that's missing.

[kendfrey]

I'm also experiencing this error on versions 0.13.0 through 0.16.0. Versions 0.12.0 and older do not have this problem.