Previous yolo-v3.ot outdated / yolo/main.rs needs update

[ralfbiedert]

First of all thank you so much for tch-rs! I've been working with it for a while now and it has been an absolute joy to use!

I was now trying to run the yolo example and downloaded the weights mentioned in yolo/main.rs, which was also recommended in #337, #334 or #293.

However, when trying to run cargo run --example yolo yolo-v3.ot cat.jpg with said file I get

Error: Internal torch error: Legacy model format is not supported on mobile.
Exception raised from deserialize at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\torch\csrc\jit\serialization\import.cpp:273 (most recent call first):
00007FFBE7F7DA2200007FFBE7F7D9C0 c10.dll!c10::Error::Error [<unknown file> @ <unknown line number>]
00007FFBE7F7D43E00007FFBE7F7D3F0 c10.dll!c10::detail::torchCheckFail [<unknown file> @ <unknown line number>]
00007FFB5593344200007FFB55930CB0 torch_cpu.dll!torch::jit::CallStackDebugInfoUnpickler::unpickle [<unknown file> @ <unknown line number>]
00007FFB5593557D00007FFB55935400 torch_cpu.dll!torch::jit::load [<unknown file> @ <unknown line number>]
00007FFB559351ED00007FFB559350E0 torch_cpu.dll!torch::jit::load [<unknown file> @ <unknown line number>]
00007FFB5593502A00007FFB55934F80 torch_cpu.dll!torch::jit::load [<unknown file> @ <unknown line number>]
00007FF7CAE1052900007FF7CAE10490 yolo.exe!at_load_callback_with_device [C:\Users\rb\Development\Source\_thirdparty\tch-rs\torch-sys\libtch\torch_api.cpp @ 418]
00007FF7CAD848F900007FF7CAD847C0 yolo.exe!tch::wrappers::tensor::Tensor::load_multi_with_device<ref$<ref$<alloc::string::String> > > [C:\Users\rb\Development\Source\_thirdparty\tch-rs\src\wrappers\tensor.rs @ 643]
00007FF7CAD69E3E00007FF7CAD69DF0 yolo.exe!tch::nn::var_store::VarStore::load<ref$<alloc::string::String> > [C:\Users\rb\Development\Source\_thirdparty\tch-rs\src\nn\var_store.rs @ 177]
00007FF7CAD6C46F00007FF7CAD6C170 yolo.exe!yolo::main [C:\Users\rb\Development\Source\_thirdparty\tch-rs\examples\yolo\main.rs @ 119]
00007FF7CAD7F07B00007FF7CAD7F070 yolo.exe!core::ops::function::FnOnce::call_once<enum2$<core::result::Result<tuple$<>,anyhow::Error> > (*)(),tuple$<> > [/rustc/95a992a68694d8bf3959bd2c0ac27ce9e9208b59\library\core\src\ops\function.rs @ 248]
00007FF7CAD6288E00007FF7CAD62880 yolo.exe!std::sys_common::backtrace::__rust_begin_short_backtrace<enum2$<core::result::Result<tuple$<>,anyhow::Error> > (*)(),enum2$<core::result::Result<tuple$<>,anyhow::Error> > > [/rustc/95a992a68694d8bf3959bd2c0ac27ce9e9208b59\library\std\src\sys_common\backtrace.rs @ 122]
00007FF7CAD8729100007FF7CAD87280 yolo.exe!std::rt::lang_start::closure$0<enum2$<core::result::Result<tuple$<>,anyhow::Error> > > [/rustc/95a992a68694d8bf3959bd2c0ac27ce9e9208b59\library\std\src\rt.rs @ 166]
00007FF7CADEBC7E00007FF7CADEBBC0 yolo.exe!std::rt::lang_start_internal [/rustc/95a992a68694d8bf3959bd2c0ac27ce9e9208b59/library\std\src\rt.rs @ 148]
00007FF7CAD8726A00007FF7CAD87230 yolo.exe!std::rt::lang_start<enum2$<core::result::Result<tuple$<>,anyhow::Error> > > [/rustc/95a992a68694d8bf3959bd2c0ac27ce9e9208b59\library\std\src\rt.rs @ 165]
00007FF7CAD6D29C00007FF7CAD6D280 yolo.exe!main [<unknown file> @ <unknown line number>]
00007FF7CAF588B400007FF7CAF587A8 yolo.exe!__scrt_common_main_seh [D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\exe_common.inl @ 288]
00007FFC3D7054E000007FFC3D7054D0 KERNEL32.DLL!BaseThreadInitThunk [<unknown file> @ <unknown line number>]
00007FFC3F24485B00007FFC3F244830 ntdll.dll!RtlUserThreadStart [<unknown file> @ <unknown line number>]

I'll try to follow this here next, but I think the example needs an updated description in any case.

[LaurentMazare]

Thanks for reporting this. I've updated the weights from the link you mentioned so this should hopefully be fixed now. I've also added a bit more details in the yolo example on how the weights could be generated.

[ralfbiedert]

Thanks, I can try tomorrow and will report back!

When diving into this I've also encountered a few questions for which I've opened a new ticket here #549.

[ralfbiedert]

Worked, thanks a lot!