Need help to get mnist.py working

[Escain]

Hi,

I am trying to get this mnist.py example working, but so far it's crashing.

I created the following BASH script to make the full install and build process:

#!/bin/bash

PROJDIR=$(pwd)/pytorch_dlprim

# Clone pytorch_dlprim
git clone --recurse-submodules https://github.com/artyom-beilis/pytorch_dlprim.git "$PROJDIR"
cd "$PROJDIR"

# Get python version (3.11 in my case)
python_version=$(python3 --version 2>&1)  # Captures the output of `python --version`
python_major_minor=$(echo "$python_version" | cut -d ' ' -f 2)  # Extracts the second field, which is the version number
python_major=$(echo "$python_major_minor" | cut -d '.' -f 1)  # Extracts the major version
python_minor=$(echo "$python_major_minor" | cut -d '.' -f 2)  # Extracts the minor version

# Create python environment
python3 -m venv env
VIRTUAL_ENV="$(pwd)/env"
source $VIRTUAL_ENV/bin/activate

# Download torchvision CPU
python -m pip install --force-reinstall --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu

# Build dlprim
mkdir build
cd build
cmake -DCMAKE_PREFIX_PATH=$VIRTUAL_ENV/lib/python$python_major.$pythonminor/site-packages/torch/share/cmake/Torch ..
make -j15
cd ..

python mnist.py --device ocl:0

exit

I also try to get the pytorch 1.13 as suggested, but it seems that this version cannot be downloaded anymore?
pip install torch==1.13.1+cpu torchvision==0.14.1+cpu torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cpu Fails with requering torchaudo 2+ and torchvision 0.15+. This consequently requirs torch v2+

My computer is:

• Radeon 7900 with Mesa driver
• 16Gb ram
• Threadripper
• Linux 6.1

mnist.py crash

using GDB, I get the following back-trace:

Thread 1 "python" received signal SIGSEGV, Segmentation fault.
__memcpy_avx_unaligned () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:228
228     ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S: No such file or directory.
(gdb) bt
#0  __memcpy_avx_unaligned () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:228
#1  0x00007fffde73e618 in std::basic_streambuf<char, std::char_traits<char> >::xsputn(char const*, long) () at /lib/x86_64-linux-gnu/libstdc++.so.6
#2  0x00007fffde73088d in std::basic_ostream<char, std::char_traits<char> >& std::__ostream_insert<char, std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&, char const*, long) () at /lib/x86_64-linux-gnu/libstdc++.so.6
#3  0x00007fffaa15e9e7 in c10::Dispatcher::registerFallback(c10::DispatchKey, c10::KernelFunction, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) () at /lib/x86_64-linux-gnu/libtorch_cpu.so.1.13
#4  0x00007fffaa19e2cc in torch::Library::_fallback(torch::CppFunction&&) & () at /lib/x86_64-linux-gnu/libtorch_cpu.so.1.13
#5  0x00007fffa9f23075 in  () at /lib/x86_64-linux-gnu/libtorch_cpu.so.1.13
#6  0x00007fffa9f29619 in  () at /lib/x86_64-linux-gnu/libtorch_cpu.so.1.13
#7  0x00007fffa9ec77ab in  () at /lib/x86_64-linux-gnu/libtorch_cpu.so.1.13
#8  0x00007ffff7fcf8be in call_init (env=0x34d61b0, argv=0x7fffffffdb68, argc=4, l=<optimized out>) at ./elf/dl-init.c:90
#9  call_init (l=<optimized out>, argc=4, argv=0x7fffffffdb68, env=0x34d61b0) at ./elf/dl-init.c:27
#10 0x00007ffff7fcf9a4 in _dl_init (main_map=0x6011350, argc=4, argv=0x7fffffffdb68, env=0x34d61b0) at ./elf/dl-init.c:137
#11 0x00007ffff7dd8024 in __GI__dl_catch_exception (exception=<optimized out>, operate=<optimized out>, args=<optimized out>) at ./elf/dl-error-skeleton.c:182
#12 0x00007ffff7fd609e in dl_open_worker (a=a@entry=0x7fffffffcfd0) at ./elf/dl-open.c:808
#13 0x00007ffff7dd7fca in __GI__dl_catch_exception (exception=<optimized out>, operate=<optimized out>, args=<optimized out>) at ./elf/dl-error-skeleton.c:208
#14 0x00007ffff7fd6438 in _dl_open
    (file=0x7fffb06fa510 "/home/...../test_myxtral/pytorch_dlprim/build/libpt_ocl.so", mode=<optimized out>, caller_dlopen=0x7ffff7b05ae9, nsid=<optimized out>, argc=4, argv=0x7fffffffdb68, env=0x34d61b0)
    at ./elf/dl-open.c:884
#15 0x00007ffff7d0e438 in dlopen_doit (a=a@entry=0x7fffffffd240) at ./dlfcn/dlopen.c:56
#16 0x00007ffff7dd7fca in __GI__dl_catch_exception (exception=exception@entry=0x7fffffffd1a0, operate=<optimized out>, args=<optimized out>) at ./elf/dl-error-skeleton.c:208
#17 0x00007ffff7dd807f in __GI__dl_catch_error (objname=0x7fffffffd1f8, errstring=0x7fffffffd200, mallocedp=0x7fffffffd1f7, operate=<optimized out>, args=<optimized out>) at ./elf/dl-error-skeleton.c:227
#18 0x00007ffff7d0df27 in _dlerror_run (operate=operate@entry=0x7ffff7d0e3e0 <dlopen_doit>, args=args@entry=0x7fffffffd240) at ./dlfcn/dlerror.c:138
#19 0x00007ffff7d0e4e9 in dlopen_implementation (dl_caller=<optimized out>, mode=<optimized out>, file=<optimized out>) at ./dlfcn/dlopen.c:71
#20 ___dlopen (file=<optimized out>, mode=<optimized out>) at ./dlfcn/dlopen.c:81
#21 0x00007ffff7b05ae9 in  () at /usr/lib/python3.11/lib-dynload/_ctypes.cpython-311-x86_64-linux-gnu.so
#22 0x0000000000547ae8 in  ()
#23 0x0000000000517eb3 in _PyObject_MakeTpCall ()
#24 0x000000000052b940 in _PyEval_EvalFrameDefault ()
#25 0x000000000051f067 in _PyObject_FastCallDictTstate ()
#26 0x0000000000565973 in  ()
#27 0x0000000000517e55 in _PyObject_MakeTpCall ()
#28 0x000000000052b940 in _PyEval_EvalFrameDefault ()
#29 0x000000000052360b in PyEval_EvalCode ()
#30 0x0000000000647497 in  ()
#31 0x0000000000644d4f in  ()
#32 0x0000000000651010 in  ()
#33 0x0000000000650d5b in _PyRun_SimpleFileObject ()
#34 0x0000000000650b84 in _PyRun_AnyFileObject ()
#35 0x000000000064f90f in Py_RunMain ()
#36 0x00000000006275c7 in Py_BytesMain ()
#37 0x00007ffff7cb024a in __libc_start_call_main (main=main@entry=0x627530, argc=argc@entry=4, argv=argv@entry=0x7fffffffdb68) at ../sysdeps/nptl/libc_start_call_main.h:58
#38 0x00007ffff7cb0305 in __libc_start_main_impl (main=0x627530, argc=4, argv=0x7fffffffdb68, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffdb58) at ../csu/libc-start.c:360
#39 0x0000000000627461 in _start ()

Any suggestion?