perrying / ssn-pytorch Goto Github PK

View Code? Open in Web Editor NEW

117.0 117.0 22.0 605 KB

PyTorch implementation of Superpixel Sampling Networks

License: MIT License

Python 100.00%

ssn-pytorch's People

Contributors

Stargazers

Watchers

ssn-pytorch's Issues

the reconstructed features may have been calculated incorrectly

ssn-pytorch/lib/utils/loss.py

Lines 49 to 53 in 25c69d3

    
           if hard_assignment is None: 
        
               # (B, n_spixels, n_pixels) -> (B, n_pixels, n_spixels) 
        
               permuted_assignment = assignment.permute(0, 2, 1).contiguous() 
        
               # matrix product between (n_pixels, n_spixels) and (n_spixels, channels) 
        
               reconstructed_labels = torch.bmm(permuted_assignment, spixel_mean)

I have a question with line 53 in lib/utils/loss.py. According the paper 'superpixel sampling networks', the reconstructed pixel features are , and means the row-normalized association matrix Q. however, the code in line 53 seems not using row-normalized Q but the Q itself to compute the reconstructed label.

why compile cuda source

why compile cuda source
something wrong in the "from lib.ssn.ssb import sparse_ssn_iter"?why?
should i compile the ssn firstly?

I cannot debug your code successfully

First of all, thank you very much for sharing the code. After many days of debugging your code, I still cannot run it successfully. My operating environment is cuda10.1 + torch 1.4.0, and my current debugging error is as follows:
RecursionError: maximum recursion depth exceeded in comparison
This is a problem that I cannot solve.
Could you please write a detailed operating instruction and provide a detailed operating environment?
If you are willing to email me, I will be very grateful to you!
email: [email protected]

Hello, I wonder how to set the number of superpixels in the results.

pair_wise_distance文件里，cpp_sources为空吗，我运行的时候出现问题：'cp1' codec can't decode bytes in position 0--1: No mapping for the Unicode character exists in the target code page，麻烦作者可以帮我看看这是怎么回事吗

pair_wise_distance_cuda = load_inline(
"pair_wise_distance", cpp_sources='', cuda_sources=source
)

Pretrained weights

Do you or anyone else have pre trained weights for this?

THC.h : no such file or directory

Traceback (most recent call last):
  File "/home/ppavia/.local/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1740, in _run_ninja_build
    subprocess.run(
  File "/usr/lib/python3.8/subprocess.py", line 516, in run
    raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "train.py", line 9, in <module>
    from model import SSNModel
  File "/data3/ppavia/ssn-pytorch/model.py", line 4, in <module>
    from lib.ssn.ssn import ssn_iter, sparse_ssn_iter
  File "/data3/ppavia/ssn-pytorch/lib/ssn/ssn.py", line 4, in <module>
    from .pair_wise_distance import PairwiseDistFunction
  File "/data3/ppavia/ssn-pytorch/lib/ssn/pair_wise_distance.py", line 8, in <module>
    pair_wise_distance_cuda = load_inline(
  File "/home/ppavia/.local/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1293, in load_inline
    return _jit_compile(
  File "/home/ppavia/.local/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1357, in _jit_compile
    _write_ninja_file_and_build_library(
  File "/home/ppavia/.local/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1469, in _write_ninja_file_and_build_library
    _run_ninja_build(
  File "/home/ppavia/.local/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1756, in _run_ninja_build
    raise RuntimeError(message) from e
RuntimeError: Error building extension 'pair_wise_distance': [1/3] /usr/local/cuda/bin/nvcc  -DTORCH_EXTENSION_NAME=pair_wise_distance -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/TH -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_70,code=compute_70 -gencode=arch=compute_70,code=sm_70 --compiler-options '-fPIC' -std=c++14 -c /home/ppavia/.cache/torch_extensions/py38_cu102/pair_wise_distance/cuda.cu -o cuda.cuda.o 
FAILED: cuda.cuda.o 
/usr/local/cuda/bin/nvcc  -DTORCH_EXTENSION_NAME=pair_wise_distance -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/TH -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_70,code=compute_70 -gencode=arch=compute_70,code=sm_70 --compiler-options '-fPIC' -std=c++14 -c /home/ppavia/.cache/torch_extensions/py38_cu102/pair_wise_distance/cuda.cu -o cuda.cuda.o 
/home/ppavia/.cache/torch_extensions/py38_cu102/pair_wise_distance/cuda.cu:17:10: fatal error: THC/THC.h: Aucun fichier ou dossier de ce type
   17 | #include <THC/THC.h>
      |          ^~~~~~~~~~~
compilation terminated.
[2/3] c++ -MMD -MF main.o.d -DTORCH_EXTENSION_NAME=pair_wise_distance -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/TH -isystem /home/ppavia/.local/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++14 -c /home/ppavia/.cache/torch_extensions/py38_cu102/pair_wise_distance/main.cpp -o main.o 
ninja: build stopped: subcommand failed.

Run without GPU

Is there a way to run the the pair_wise_distance without gpu or cuda? Thanks!

CUDA out of memory

RuntimeError: CUDA out of memory. Tried to allocate 7.74 GiB (GPU 0; 23.69 GiB total capacity; 18.60 GiB already allocated; 2.78 GiB free; 19.05 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Thanks very much from your work, and I need some assistance to train my own dataset, I change dimension 50 to 3 in convert_label function, but when starting training, it still reports CUDA out of memory error, I don't kown how to fix it

RuntimeError: CUDA error: an illegal memory access was encountered

there is a runtimen error of cuda when runs the following line code with pytorch 1.6 and cudatoolkit 10.2

ssn-pytorch/lib/ssn/ssn.py

Line 151 in 25c69d3

    
           sparse_abs_affinity = torch.sparse_coo_tensor(abs_indices[:, mask], reshaped_affinity_matrix[mask])

Connected to pydev debugger (build 202.6397.98)
compile cuda source of 'pair_wise_distance' function...
NOTE: if you avoid this process, you make .cu file and compile it following https://pytorch.org/tutorials/advanced/cpp_extension.html
done
Traceback (most recent call last):
File "/home/vge/pycharm-community-2020.2/plugins/python-ce/helpers/pydev/pydevd.py", line 1448, in _exec
pydev_imports.execfile(file, globals, locals) # execute the script
File "/home/vge/pycharm-community-2020.2/plugins/python-ce/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "/home/vge/yhj/projects/ssn-pytorch/train.py", line 156, in
train(args)
File "/home/vge/yhj/projects/ssn-pytorch/train.py", line 117, in train
metric = update_param(data, model, optimizer, cfg.compactness, cfg.color_scale, cfg.pos_scale, device)
File "/home/vge/yhj/projects/ssn-pytorch/train.py", line 79, in update_param
Q, H, feat = model(inputs)
File "/home/vge/anaconda3/envs/pytorch_yhj/lib/python3.8/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/vge/yhj/projects/ssn-pytorch/model.py", line 51, in forward
return ssn_iter(pixel_f, self.nspix, self.n_iter)
File "/home/vge/yhj/projects/ssn-pytorch/lib/ssn/ssn.py", line 151, in ssn_iter
sparse_abs_affinity = torch.sparse_coo_tensor(abs_indices[:, mask], reshaped_affinity_matrix[mask])
RuntimeError: CUDA error: an illegal memory access was encountered
python-BaseException
Backend TkAgg is interactive backend. Turning interactive mode on.
THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1595629395347/work/aten/src/THC/THCCachingHostAllocator.cpp line=278 error=700 : an illegal memory access was encountered

Process finished with exit code 130 (interrupted by signal 2: SIGINT)

Missing dependency

This requires the ninja package, with unspecified version requirements, and that's missing from the readme.

perrying / ssn-pytorch Goto Github PK

ssn-pytorch's People

Contributors

Stargazers

Watchers

Forkers

ssn-pytorch's Issues

the reconstructed features may have been calculated incorrectly

why compile cuda source

I cannot debug your code successfully

Hello, I wonder how to set the number of superpixels in the results.

pair_wise_distance文件里，cpp_sources为空吗，我运行的时候出现问题：'cp1' codec can't decode bytes in position 0--1: No mapping for the Unicode character exists in the target code page，麻烦作者可以帮我看看这是怎么回事吗

Pretrained weights

THC.h : no such file or directory

Run without GPU

CUDA out of memory

RuntimeError: CUDA error: an illegal memory access was encountered

Missing dependency

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	if hard_assignment is None:
	# (B, n_spixels, n_pixels) -> (B, n_pixels, n_spixels)
	permuted_assignment = assignment.permute(0, 2, 1).contiguous()
	# matrix product between (n_pixels, n_spixels) and (n_spixels, channels)
	reconstructed_labels = torch.bmm(permuted_assignment, spixel_mean)