jdnie / addernetcuda Goto Github PK

I followed the instructions in the readme file to run the code and getting the following error. Am I doing something wrong?

File "/content/utils.py", line 39, in accuracy
correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

I have not made any changes to the code.

Hello, I have a question. It seems that you set your code running on just one gpu, and how can I run via multi gpu?

Hi @jdnie

Your code is impressive. However, when I try to use the cuda and run "python setup.py install", there are a lot of error.

What is the reason? I follow your instruction to use pytorch 1.2.0 and torchvision 0.4.0.

I have download the whole files AdderNetCuda, and then I put it on cloud platform.
cd unoptimized
python setup.py install
python test.py

Question is :

• I want to know the solution to the problem "Unable to import CUDA unoptimized kernels"

The environment is :
NVIDIA GeForce RTX 3080, 1

The version of torch and torchvision are :
torch 1.10.0+cu111
torchvision 0.11.0+cu111

However, I cannot get the results of method using cuda. The program gives me

• Unable to import CUDA unoptimized kernels
  x:=================
  tensor([[[[0.1350, 0.9071, 0.3186],
  [0.4736, 0.1879, 0.0715],
  [0.7890, 0.5512, 0.1568]]]], device='cuda:0')
  w:=================
  tensor([[[[0.9958]]]], device='cuda:0')
  grad_y:=================
  tensor([[[[0.9757, 0.5837],
  [0.2281, 0.6243]]]], device='cuda:0')
  Traceback (most recent call last):
  File "test.py", line 310, in
  conv_input_test()
  File "test.py", line 287, in conv_input_test
  grad_x = conv_input(grad_y, x, w, [stride,], [padding,])
  File "test.py", line 86, in conv_input
  unoptimized_cuda.UNOPTIMIZED_CONV_INPUT(grad_y, x, w, grad_x, strides, paddings)
  NameError: name 'unoptimized_cuda' is not defined

The log of install unoptimized is :

• ......
• creating build/lib.linux-x86_64-cpython-38
  x86_64-linux-gnu-g++ -pthread -shared -Wl,-O1 -Wl,-Bsymbolic-functions -Wl,-Bsymbolic-functions -Wl,-z,relro -g -fwrapv -O2 build/temp.linux-x86_64-cpython-38/unoptimized.o build/temp.linux-x86_64-cpython-38/unoptimized_cuda.o -L/usr/local/lib/python3.8/dist-packages/torch/lib -L/usr/local/cuda/lib64 -L/usr/lib -lc10 -ltorch -ltorch_cpu -ltorch_python -lcudart -lc10_cuda -ltorch_cuda_cu -ltorch_cuda_cpp -o build/lib.linux-x86_64-cpython-38/unoptimized_cuda.cpython-38-x86_64-linux-gnu.so
  creating build/bdist.linux-x86_64
  creating build/bdist.linux-x86_64/egg
  copying build/lib.linux-x86_64-cpython-38/unoptimized_cuda.cpython-38-x86_64-linux-gnu.so -> build/bdist.linux-x86_64/egg
  creating stub loader for unoptimized_cuda.cpython-38-x86_64-linux-gnu.so
  byte-compiling build/bdist.linux-x86_64/egg/unoptimized_cuda.py to unoptimized_cuda.cpython-38.pyc
  creating build/bdist.linux-x86_64/egg/EGG-INFO
  copying unoptimized_cuda.egg-info/PKG-INFO -> build/bdist.linux-x86_64/egg/EGG-INFO
  copying unoptimized_cuda.egg-info/SOURCES.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
  copying unoptimized_cuda.egg-info/dependency_links.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
  copying unoptimized_cuda.egg-info/top_level.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
  writing build/bdist.linux-x86_64/egg/EGG-INFO/native_libs.txt
  zip_safe flag not set; analyzing archive contents...
  pycache.unoptimized_cuda.cpython-38: module references file
  creating dist
  creating 'dist/unoptimized_cuda-0.0.0-py3.8-linux-x86_64.egg' and adding 'build/bdist.linux-x86_64/egg' to it
  removing 'build/bdist.linux-x86_64/egg' (and everything under it)
  Processing unoptimized_cuda-0.0.0-py3.8-linux-x86_64.egg
  creating /usr/lib/python3.8/site-packages/unoptimized_cuda-0.0.0-py3.8-linux-x86_64.egg
  Extracting unoptimized_cuda-0.0.0-py3.8-linux-x86_64.egg to /usr/lib/python3.8/site-packages
  Adding unoptimized-cuda 0.0.0 to easy-install.pth file

Installed /usr/lib/python3.8/site-packages/unoptimized_cuda-0.0.0-py3.8-linux-x86_64.egg
Processing dependencies for unoptimized-cuda==0.0.0
Finished processing dependencies for unoptimized-cuda==0.0.0

used your training code, the accuracy is about 50%，but the paper's accuracy has achieved 90%,could you please tell me why?