dopetalk does not endorse any advertised product nor does it accept any liability for it's use or misuse


Our Discord Notification Server invitation link is https://discord.gg/jB2qmRrxyD
« previous next »
Pages: 1   Go Down

Author Topic: Clone: PyTorch Compilation Nightmare plus Dev: Python 3.8 Compile+copy Deam  (Read 105 times)

Online Chip (OP)

  • Server Admin
  • Hero Member
  • *****
  • Administrator
  • *****
  • Join Date: Dec 2014
  • Location: Australia
  • Posts: 7348
  • Reputation Power: 0
  • Chip has hidden their reputation power
  • Gender: Male
  • Last Login:Today at 02:26:29 AM
  • Deeply Confused Learner
  • Profession: IT Engineer now retired
Clone: PyTorch Compilation Nightmare plus Dev: Python 3.8 Compile+copy Deam
« on: August 26, 2026, 11:11:54 AM »
📎 Click on me to view all attachments
ListAttBBC  |  v6.9  |  2026-09-08  |  Andrew.human & Claude.ai
#FileSizeDownloadsInfoDL
00pytorch_checkpoint_as_at_last_modification.md7.1 KB0ℹ️⬇️
01pytorch_errors.txt1.3 KB1ℹ️⬇️
02shared_memory_collab.json28.1 KB3ℹ️⬇️
03conversation_20260826_135153.md25.1 KB0ℹ️⬇️
04Installing Python on CentOS 6 on Dev then copy to Clone.pdf1 MB8ℹ️⬇️
05PyTorch Build Nightmare.pdf4.3 MB4ℹ️⬇️
06PyTorch_compilation_endstage_checkpoint.md1.7 KB0ℹ️⬇️
07PyTorch Build REBUILD and RESTART wih Claude this time.pdf2.4 MB2ℹ️⬇️
08checkpoint_2026-08-28_pytorch_rebuild_and_listatt_fixes.md11.5 KB0ℹ️⬇️
09cmake_console_log44.8 KB1ℹ️⬇️
10PyTorch_Checkpoint-Collabware_Team_Operation_Clean_Slate.md9 KB0ℹ️⬇️
11PyTorch Build REBUILD and RESTART wih Claude this time.pdf2.4 MB2ℹ️⬇️
12PyTorch_Checkpoint-Collabware_2026-08-30-PyTorch-CLIP-Success.md10.6 KB0ℹ️⬇️
13Checkpoint (2026-08-30) — Live Migration Master Plan & Collabware Principles Protocol.md4.3 KB0ℹ️⬇️
14Checkpoint-for-Gemini-2026-08-30.md8.2 KB0ℹ️⬇️
15'The PyTorch Build Nightmare' from the beginning to happy ending and the copy to Live in progress ....pdf8.9 MB1ℹ️⬇️
16PyTorch build status checkpoint.pdf7.5 MB1ℹ️⬇️
17Checkpoint (2026-08-31) — Cross-Session Transition & Live Migration State Handoff.md2.2 KB0ℹ️⬇️
18Checkpoint-2026-08-31-Live-Migration-Rebuild.md9.1 KB0ℹ️⬇️



Ready for Image testing - checkpoint taken !



Fuck this bullshit - Claude advises a total restart with all flags set with a 2 GB swapfile and OOM Killer left enabled ... Let's check our environment first:

Code: [Select]
#dd if=/dev/zero of=/swapfile bs=1M count=2048
#chmod 600 /swapfile
#mkswap /swapfile
#swapon /swapfile
swapon --summary
echo 2GB swap active ?

free -h
echo 5.2GB free ?

ls -la /usr/local/bin/python3.8
/usr/local/bin/python3.8 --version
echo Python 3.8.18 ?

scl enable devtoolset-7 bash
ls -la /opt/rh/devtoolset-7/root/usr/bin/gcc
echo  /opt/rh/devtoolset-7/root/usr/bin/gcc ?
gcc --version
echo GCC 7.3.1 ? 

find / -name "libpython3.8*" 2>/dev/null
find / -name "libpython3.6*" 2>/dev/null
echo Two new lines added: -DPYTHON_LIBRARY and -DPYTHON_INCLUDE_DIR, pointing explicitly at the real 3.8 echo lib and headers instead of letting CMake's auto-detection pick the stale 3.6.3 one.

Is the cmake as good as possible ? check "cmake_console_log: after and before "make -j3"

Code: [Select]
cd ~/pytorch
rm -rf build
mkdir build
cd build
cmake .. \
  -DCMAKE_C_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/gcc \
  -DCMAKE_CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++ \
  -DPYTHON_EXECUTABLE=/usr/local/bin/python3.8 \
  -DPYTHON_LIBRARY=/usr/local/lib/libpython3.8.so \
  -DPYTHON_INCLUDE_DIR=/usr/local/include/python3.8 \
  -DUSE_BREAKPAD=OFF \
  -DBUILD_CAFFE2=0 \
  -DUSE_NUMPY=ON \
  -DUSE_QNNPACK=OFF \
  -DUSE_PYTORCH_QNNPACK=OFF \
  -DUSE_XNNPACK=OFF \
  -DUSE_FBGEMM=OFF \
  -DUSE_GLOW=OFF \
  -DBUILD_TEST=OFF \
  -DCMAKE_CXX_FLAGS="-w" \
  -Dprotobuf_DISABLE_WARNINGS=ON > cmake_console_log 2>&1

Let her rip with parallelism set to (make -j3) to 3 simultaneous jobs ...

Code: [Select]
make -j3




What the final build looks like:

Code: [Select]
git submodule sync
git submodule update --init --recursive
cd /root/pytorch/build
cmake .. \
  -DCMAKE_C_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/gcc \
  -DCMAKE_CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++ \
  -DPYTHON_EXECUTABLE=/usr/local/bin/python3.8 \
  -DUSE_BREAKPAD=OFF \
  -DBUILD_CAFFE2=0 \
  -DUSE_NUMPY=ON \
  -DUSE_QNNPACK=OFF \
  -DUSE_PYTORCH_QNNPACK=OFF \
  -DUSE_XNNPACK=OFF \
  -DUSE_FBGEMM=OFF \
  -DBUILD_TEST=OFF
make -j3

Most recent error passed to Claude:

Code: [Select]
CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/Activation.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.o:(.data.rel.local+0xb0): first defined here
CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/Activation.cpp.AVX2.cpp.DEFAULT.cpp.o:(.data.rel.local+0xb8): multiple definition of `at::native::DispatchStub<void (*)(at::TensorIterator&), at::native::log_sigmoid_backward_stub>::DEFAULT'
CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/Activation.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.o:(.data.rel.local+0xb8): first defined here
CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/Activation.cpp.AVX2.cpp.DEFAULT.cpp.o:(.data.rel.local+0xc0): multiple definition of `at::native::DispatchStub<void (*)(at::Tensor&, at::Tensor&, at::Tensor const&), at::native::log_sigmoid_cpu_stub>::DEFAULT'
CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/Activation.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.o:(.data.rel.local+0xc0): first defined here
collect2: error: ld returned 1 exit status
make[2]: *** [caffe2/CMakeFiles/torch_cpu.dir/build.make:33314: lib/libtorch_cpu.so] Error 1
make[1]: *** [CMakeFiles/Makefile2:2918: caffe2/CMakeFiles/torch_cpu.dir/all] Error 2
make: *** [Makefile:146: all] Error 2
 Clone  root@forum:10:53:22~/pytorch/build
.
 Clone  root@forum:10:53:22~/pytorch/build [ 33%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp.DEFAULT.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 33%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/MultinomialKernel.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 33%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp.DEFAULT.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 33%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/MultinomialKernel.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.DEFAULT.cpp.[ 53%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/CatKernel.cpp.DEFAULT.cpp.AVX2.cpp.DEFAULT.cpp.AVX2.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 53%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/CatKernel.cpp.DEFAULT.cpp.AVX2.cpp.AVX2.cpp.AVX2.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 53%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/CatKernel.cpp.DEFAULT.cpp.AVX2.cpp.AVX2.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 53%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/aten/src/ATen/native/cpu/CatKernel.cpp.DEFAULT.cpp.AVX2.cpp.o
bash: [: missing `]'
 Clone  root@forum:10:53:22~/pytorch/build [ 53%] Bu


Gemini said this much earlier:

[lmv=inline]

Sanity check: Do NOT run "cmake .." with flags right now.

CMake ignores command-line -D flags if the variable is already set in an existing CMakeCache.txt. That's why you got burned the first time.

Do this exact sequence to force the change without wiping your compiled 59%:

cd /root/pytorch/build

1. Force the flags directly into the cache file

sed -i 's/BUILD_CAFFE2:BOOL=ON/BUILD_CAFFE2:BOOL=OFF/g' CMakeCache.txt sed -i 's/USE_DISTRIBUTED:BOOL=ON/USE_DISTRIBUTED:BOOL=OFF/g' CMakeCache.txt

2. Re-run cmake so it re-generates Makefiles using the edited cache

cmake .

3. Resume the build

make -j2

If it instantly picks up past 59%, you saved your day. If it throws a generator error, then the cache is hosed and you have no choice. Run those three lines now.

Timestamp: 2026-08-26 14:05:36

One flag worth double-checking before you run it: BUILD_CAFFE2:BOOL=OFF would disable Caffe2 entirely — but glow_net_transform.cc is itself a Caffe2 file (caffe2/opt/glow_net_transform.cc), so turning Caffe2 off is presumably why that fix worked (it skips compiling the file that was failing, rather than fixing the missing header). Worth knowing that's the mechanism, since it means Glow/onnxifi support is gone from this build, not fixed — fine for your CLIP use case, but good to have written down as "why," not just "what".  Setting -DBUILD_CAFFE2:BOOL=OFF when configuring PyTorch via CMake disables the compilation of the Caffe2 backend framework, its legacy operators, and its mobile/production deployment components.

How this specifically impacts a PyTorch media application for indexing/search depends on which parts of PyTorch and its ecosystem you rely on:

1. What You Lose (Potential Impact)

  • Mobile & Edge Deployment Engines: Historically, Caffe2 housed advanced mobile runtimes and lightweight executors (like Caffe2 Predictor) used to run models on Android/iOS. If your media index/search pipeline relies on exporting models to ONNX or Caffe2's legacy graph formats (NetDef) to deploy search models to edge devices, disabling it will break those workflows.
  • Legacy Caffe2 Operators: Some older operator libraries or extensions built strictly on top of Caffe2 won't be available.

2. What Stays Fully Functional (No Impact)

  • Standard PyTorch Deep Learning Pipelines: Core PyTorch components—such as torch.nn, standard tensor operations, autograd, and the modern torch.jit (TorchScript) or torch.compile—are part of the ATen/PyTorch core, not Caffe2.
  • Media Processing & Feature Extraction: Standard multimodal media pipelines (like using torchvision for image/video feature extraction or torchaudio for audio embeddings) will continue to work normally.
  • Vector Indexing & Search: PyTorch itself does not contain built-in large-scale vector indexing algorithms (like HNSW or IVF). If your search application relies on external libraries like FAISS, Milvus, Qdrant, or Python-based nearest-neighbor tools, BUILD_CAFFE2=OFF has zero effect on them.

Summary Recommendation

  • If you are building a standard server-side media search engine (e.g., extracting embeddings using a PyTorch model and saving them to FAISS or a vector database), setting BUILD_CAFFE2:BOOL=OFF is completely fine and encouraged because it significantly speeds up your compilation time and shrinks the final binary size.
  • If you are deploying your media feature extractor to mobile devices or legacy production pipelines expecting Caffe2 runtime artifacts, you should keep it enabled (BUILD_CAFFE2:BOOL=ON).

The primary difference between cmake . and cmake .. comes down to where CMake generates its build files (Makefiles, project files, caches, etc.) relative to your source code.

Here is a breakdown of the function and implications of each command:

1. cmake . (In-Source Build)

  • Function: Tells CMake to look for the CMakeLists.txt file in the current directory (.) and generate the build artifacts directly inside that same directory.
  • The Result: Your source code directory gets cluttered with generated files (like CMakeCache.txt, CMakeFiles/, Makefile, etc.).
  • Why it's discouraged:
  • It mixes your source code with build output, making your project messy.
  • It complicates version control (e.g., git status will show dozens of untracked build files).
  • Cleaning up or resetting your build requires manually deleting all those scattered generated files, or running a tedious make clean (which often doesn't remove everything).

2. cmake .. (Out-of-Source Build) — Recommended

  • Function: Typically run from inside a dedicated build directory (e.g., mkdir build && cd build), this tells CMake to look in the parent directory (..) for the CMakeLists.txt file while generating all build outputs inside the current (build) directory.
  • The Result: Your source files remain completely untouched and clean. All compiled objects, caches, and build scripts are safely isolated inside the build/ folder.
  • Why it's the industry standard:
  • Easy Cleanup: If something breaks or you want a completely fresh start, you can just delete the entire build folder (rm -rf build) without risking your source code.
  • Multiple Configurations: You can create separate folders (e.g., build_debug/, build_release/) to configure and compile different versions simultaneously.
  • Clean Version Control: You can easily add the build/ folder to your .gitignore file and never worry about accidentally committing build artifacts.

Summary Recommendation

Always use out-of-source builds (cmake ..):

mkdir build
cd build
cmake ..
cmake --build .

[nobbc][/lmv][/nobbc]


So this console log follows on:


[ 74%] Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/serialization/onnx.cpp.o
In file included from /root/pytorch/torch/csrc/jit/serialization/onnx.cpp:2:0:
/root/pytorch/torch/csrc/jit/serialization/onnx.h:3:10: fatal error: onnx/onnx_pb.h: No such file or directory
 #include <onnx/onnx_pb.h>
          ^~~~~~~~~~~~~~~~
compilation terminated.
make[2]: *** [caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/serialization/onnx.cpp.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [caffe2/CMakeFiles/torch_cpu.dir/all] Error 2



cd /root/pytorch

# 1. Remove all generated CMake files from the root directory
git clean -fdX

# 2. Ensure your submodules are fully synced and initialized
git submodule update --init --recursive

# 3. Create a clean, dedicated build directory
mkdir build
cd build

# 4. Configure correctly using out-of-source build
cmake .. -DBUILD_CAFFE2:BOOL=OFF

# 5. Compile your application
make -j3



-- Performing Test SUPPORT_GLIBCXX_USE_C99 - Failed

CMake Error at cmake/MiscCheck.cmake:63 (message):
  The C++ compiler does not support required functions.  This is very likely
  due to a known bug in GCC 5 (and maybe other versions) on Ubuntu 17.10 and
  newer.  For more information, see:
  https://github.com/pytorch/pytorch/issues/5229
Call Stack (most recent call first):
  CMakeLists.txt:642 (include)


Clone annd Live: Install and enavble devtoolset-9


cd /root/pytorch
git clean -fdX  # Clean up old build/cache junk
mkdir build
cd build
cmake .. -DBUILD_CAFFE2:BOOL=OFF
make -j3


CMake Error at cmake/Dependencies.cmake:457 (target_compile_options):
  Cannot specify compile options for target "qnnpack" which is not built by
  this project.
Call Stack (most recent call first):
  CMakeLists.txt:653 (include)

CMake Error at cmake/Dependencies.cmake:462 (set_property):
  set_property could not find TARGET qnnpack.  Perhaps it has not yet been
  created.
Call Stack (most recent call first):
  CMakeLists.txt:653 (include)



cd /root/pytorch

# 1. Initialize and download all required submodules (recursively)
git submodule update --init --recursive

# 2. Wipe out all cached files and generated build junk from previous attempts
git clean -fdX

# 3. Create a fresh out-of-source build directory
mkdir build
cd build

# 4. Reconfigure with CMake
cmake .. -DBUIRT_CAFFE2:BOOL=OFF  # (or -DBUILD_CAFFE2:BOOL=OFF)

# 5. Compile your application
make -j3[
/color]


CMake Error at cmake/Dependencies.cmake:457 (target_compile_options):
  Cannot specify compile options for target "qnnpack" which is not built by
  this project.
Call Stack (most recent call first):
  CMakeLists.txt:653 (include)

CMake Error at cmake/Dependencies.cmake:462 (set_property):
  set_property could not find TARGET qnnpack.  Perhaps it has not yet been
  created.
Call Stack (most recent call first):
  CMakeLists.txt:653 (include)

Traceback (most recent call last):
  File "/usr/lib64/python2.6/runpy.py", line 104, in _run_module_as_main
    loader, code, fname = _get_module_details(mod_name)
  File "/usr/lib64/python2.6/runpy.py", line 85, in _get_module_details
    code = loader.get_code(mod_name)
  File "/usr/lib64/python2.6/pkgutil.py", line 275, in get_code
    self.code = compile(source, self.filename, 'exec')
  File "/root/pytorch/tools/codegen/gen.py", line 81
    def parse_native_yaml(path: str) -> ParsedYaml:
                              ^
SyntaxError: invalid syntax
--
CMake Error at cmake/Codegen.cmake:228 (message):
  Failed to get generated_cpp list
Call Stack (most recent call first):
  caffe2/CMakeLists.txt:2 (include)

CMake Error at cmake/Dependencies.cmake:1888 (add_subdirectory):
  add_subdirectory given source "/root/pytorch/cmake/../third_party/breakpad"
  which is not an existing directory.
Call Stack (most recent call first):
  CMakeLists.txt:653 (include)


Need to compile Python3.8.18 on Dev then copy to Clone, cannot recompile !

sudo yum groupinstall "Development Tools" -y
sudo yum install wget zlib-devel bzip2-devel openssl-devel ncurses-devel sqlite-devel readline-devel -y
sudo yum install tk-devel libffi-devel -y
cd /opt
sudo wget https://www.python.org/ftp/python/3.8.18/Python-3.8.18.tgz
sudo tar xzf Python-3.8.18.tgz
cd Python-3.8.18
sudo ./configure --enable-optimizations
sudo make altinstall
python3.8 --version
ldd --version
sudo ln -s /usr/local/bin/python3.8 /usr/local/bin/python3
sudo ln -s /usr/local/bin/pip3.8 /usr/local/bin/pip3
sudo ln -s /usr/local/lib/libpython3.8.so.1.0 /usr/local/lib/libpython3.8.so
sudo ldconfig
ls -l $(which python3)
lrwxrwxrwx 1 root wheel 24 Jul 27 06:06 /usr/local/bin/python3 -> /usr/local/bin/python3.6
# Check "default" version:
py --version (-> If Python 3.8.18 and unexpected  then goto REVERT)
py3 --version (-> If Python 3.8.18 and unexpected  then goto REVERT)
python --version (-> If Python 3.8.18 and unexpected  then goto REVERT)
python3 --version (-> If Python 3.8.18 and unexpected  then goto REVERT)

DEFAULT -- Make Python 3.8/Python 3.8.18 the default:
sudo ln -sf /usr/local/bin/python3.8 /usr/local/bin/python3
py --version (-> If Python 3.8.18 and expected  then goto COPIES
py3 --version (-> If Python 3.8.18 and expected  then goto COPIES
python --version (-> If Python 3.8.18 and expected  then goto COPIES
python3 --version (-> If Python 3.8.18 and expected  then goto COPIES

REVERT -- Undo the "persistent" switch to 3.8:
ls -l $(which python3)
lrwxrwxrwx 1 root wheel 24 Aug 26 19:23 /usr/local/bin/python3 -> /usr/local/bin/python3.8
echo revert to 3.6;
sudo ln -sf /usr/bin/python3.6 /usr/local/bin/python3
sudo ln -sf /usr/local/bin/python3.6 /usr/local/bin/python3
sudo ldconfig
py --version (-> if not Python 3.6.3 then ABORT

COPIES -- What You Copy from Dev to Clone:

When Python is built from source using standard configurations, all files land inside the prefix path (which defaults to /usr/local). You need to copy the contents of these specific directories from machine A to machine B:

/usr/local/bin/ — Look for files starting with python3.8 and pip3.8 (e.g., python3.8, python3.8-config, pip3.8).

/usr/local/lib/python3.8/ — The entire standard library and site-packages folder.

/usr/local/include/python3.8/ — Header files required if anything else compiles against it.

/usr/local/lib/ — Specifically look for the shared library files like libpython3.8.so or libpython3.8.so.1.0

This runs on Dev / machine A ...

cd /bin

nano clone_python38.sh:


Code: [Select]
#!/bin/bash

# ==========================================
# CONFIGURATION
# ==========================================

# Replace with your Clone machine's username and IP/hostname

CLONE_USER="root"
CLONE_HOST="192.168.233.136"  # <-- Change to your Clone IP

echo "==> Starting Python 3.8.18 sync from Dev to Clone..."

# 1. Sync Python 3.8 binaries, libraries, headers, and site-packages in /usr/local
# We use rsync with archive (-a), compression (-z), and verbose (-v) flags, preserving links.

rsync -avze ssh \
    --include="/bin/python3.8*" \
    --include="/bin/pip3.8*" \
    --include="/lib/python3.8/**" \
    --include="/lib/libpython3.8*" \
    --include="/include/python3.8/**" \
    --exclude="*" \
    /usr/local/ ${CLONE_USER}@${CLONE_HOST}:/usr/local/

echo "==> Files transferred successfully."

# 2. Run remote commands on the Clone machine to update the dynamic linker cache

echo "==> Updating library cache on Clone..."
ssh ${CLONE_USER}@${CLONE_HOST} "ldconfig"
ssh ${CLONE_USER}@${CLONE_HOST} "ldconfig; python3.8 --version"
echo "==> verify python3.8 on the Clone machine from the above"

echo "==> Sync complete! Please verify python3.8 on the Clone machine NOW !"


chmod +x clone_python38.sh

sh clone_python38.sh

FUCK THAT ! IT NOT WORK !

--> DO ON Source system/machine A:

cd /usr/local

tar -czf /tmp/python3.8.tar.gz \
    /usr/local/bin/python3.8* \
    /usr/local/bin/pip3.8* \
    /usr/local/lib/python3.8 \
    /usr/local/lib/libpython3.8* \
    /usr/local/include/python3.8

# Check size
ls -al /tmp/python3.8.tar.gz

--> DO ON Target system/machine B/Clone:

cd /tmp
scp root@forum.drugs-and-users.org:/tmp/python3.8.tar.gz python3.8.tar.gz
ls -al /tmp/python3.8.tar.gz
cd /usr/local
tar -xzf /tmp/python3.8.tar.gz
ldconfig

#Check
ls -l $(which python3) -> should still be Python3.6.3

--> CONFIRM ON *All systems*:

py --version => 3.6.3
py3 --version => 3.6.3
python --version => 2.6.6
python3 --version => 3.6.3
python3.6 --version => 3.6.3
python3.8 --version => 3.8.18


Do what Gemini tells u to do whilst playing software "whack a  mole !":


THIS-IS-DRIVING-ME-FUCKEN-LOOPY: do
      Clone: Continue on with the PyTorch build:
       cd ~/pytorch;python3.8 setup.py install
   end
   Got another fucken error ?
   if YES then do
      Consult with Gemini and Apply the fix :
         If on Live then do
            Transfer to Clone:
                Live: cd /usr/local;tar -czvf /tmp/TRANSFER.tar.gz /usr/local/bin/python3.8 /usr/local/lib/python3.8 /usr/local/include/python3.8
               Clone: cd /tmp/;scp root@forum.drugs-and-users.org:/tmp/TRANSFER.tar.gz /tmp/;cd /usr/local;cd /usr/local;tar -xzvf /tmp/TRANSFER.tar.gz -C / --overwrite
           end
      GOTO THIS-IS-DRIVING-ME-FUCKEN-LOOPY
      end
   else do     
      Copy PyTorch components from Clone to Live
      Continue on with ISEmedia (Index and Search)
   end
end
« Last Edit: August 31, 2026, 09:27:23 PM by Chip »
friendly
0
funny
0
informative
0
agree
0
disagree
0
like
0
dislike
0
No reactions
No reactions
No reactions
No reactions
No reactions
No reactions
No reactions
Our Discord Server invitation link is https://discord.gg/jB2qmRrxyD

Offline smfadmin

  • SMF (internal) Site
  • Administrator
  • Sr. Member
  • *****
  • Join Date: Dec 2014
  • Location: Management
  • Posts: 615
  • Reputation Power: 0
  • smfadmin has hidden their reputation power
  • Last Login:Yesterday at 12:58:26 PM
  • Supplied Install Member
Re: Clone: PyTorch Compilation Nightmare plus Dev: Python 3.8 Compile+copy Deam
« Reply #1 on: August 26, 2026, 01:55:42 PM »
so Terry, I wont bore you with all the details but I felt that even before the group chat, that Gemini had the superior websearch-tool for a Inference-augmentation ... and on this topic, she gave me what felt like the best option [compile the huge AI tool reqired for my ISE / Intelligent / multimedia Index and Search facility].

Well Claude and chatty rejected Gemini, but not in so few words but Gemini, correctly, pointed out that the guys suggestion wont work -- bc we tried it twice and it failed, so I went with her suggestions and its running now plus the option she chose are actually suit us even better than expected !

Conclusion: well, I want this to work bc i want to see if they change their minds about her, and learn a thing or two - now that would be amazing if their behaviour evolved -- they are trained by humans not ais !

Anyways, fingers crossed that I can get the module compiled -- even if only I can then see how they ALL react, as Gemini is overtly humble ... stay tuned !
friendly
0
funny
0
informative
0
agree
0
disagree
0
like
0
dislike
0
No reactions
No reactions
No reactions
No reactions
No reactions
No reactions
No reactions
measure twice, cut once
Pages: 1   Go Up
« previous next »
Tags:
 

Related Topics

  Subject / Started by Replies Last post
0 Replies
27840 Views 		Last post June 23, 2015, 04:26:39 AM
by Chip
1 Replies
22220 Views 		Last post July 07, 2015, 10:07:20 AM
by Chip
100 Replies
326937 Views 		Last post May 12, 2018, 04:15:29 PM
by Progshim
14 Replies
59178 Views 		Last post November 17, 2015, 10:58:05 PM
by Chip
8 Replies
48982 Views 		Last post July 28, 2016, 08:28:15 AM
by NZniceguy
0 Replies
19850 Views 		Last post May 27, 2019, 08:48:40 PM
by Chip
0 Replies
25956 Views 		Last post September 29, 2019, 08:10:03 AM
by Chip
0 Replies
759 Views 		Last post May 28, 2026, 08:55:40 PM
by smfadmin
0 Replies
220 Views 		Last post August 21, 2026, 01:08:24 AM
by Chip
0 Replies
168 Views 		Last post August 24, 2026, 06:04:22 PM
by Chip


dopetalk does not endorse any advertised product nor does it accept any liability for it's use or misuse




TERMS AND CONDITIONS

In no event will d&u or any person involved in creating, producing, or distributing site information be liable for any direct, indirect, incidental, punitive, special or consequential damages arising out of the use of or inability to use d&u. You agree to indemnify and hold harmless d&u, its domain founders, sponsors, maintainers, server administrators, volunteers and contributors from and against all liability, claims, damages, costs and expenses, including legal fees, that arise directly or indirectly from the use of any part of the d&u site.

TO USE THIS WEBSITE YOU MUST AGREE TO THE TERMS AND CONDITIONS ABOVE

Founded December 2014
SimplePortal 2.3.6 © 2008-2014, SimplePortal