nanocad-lab / 3pxnet Goto Github PK

Getting error Makefile:24: *** missing separator. Stop.

when calling make prereq from the 3pxnet-training directory

Following the readme of the inference directory I have run the command make validation and left the compiler options at the top of the makefile as default and the command has run fine. However when I run the python scripts/runValidation.py command I am met with the following error for every test

Running: ./validation -f -i 100
'.' is not recognized as an internal or external command,
operable program or batch file.

After calling make MNIST=1 convert from the training directory I receive the error:

File "C:\Users\andre\AppData\Local\Programs\Python\Python39\lib\site-packages\nnef_tools\conversion\converter.py", line 139, in call
message = self._check_conditions(op, transform)
File "C:\Users\andre\AppData\Local\Programs\Python\Python39\lib\site-packages\nnef_tools\conversion\converter.py", line 185, in _check_conditions
op_attribs = self._add_default_attribs(op.attribs, transform.defaults, op_inputs, op_outputs, op.type, op.name)
AttributeError: 'NoneType' object has no attribute 'defaults'
make: *** [Makefile:17: convert] Error 1

I have moved the convert.py file to the training directory but after doing this I received the error:

python convert.py --input-format onnx --output-format nnef --input-mode "training_data/FC_Large.onnx" --output-mode "../FC_Large.nnef"
Traceback (most recent call last):
File "C:\Users\andre\source\repos\3pxnet\3pxnet-training\convert.py", line 15, in
from .conversion import *
ImportError: attempted relative import with no known parent package
make: *** [Makefile:17: convert] Error 1

I solved this by removing the relative imports and instead using the solution found at How to import functions from other projects in Python? such that the top of the file now is:

from os import path
import sys
sys.path.append(path.abspath('Users/andre/source/repos/NNEF-Tools'))

from nnef_tools.conversion import *
from nnef_tools.model import utils

I understand this may be an issue to do with the nnef-tools and not with your repository and if so please close this error.

#Issue with training CIFAR 3pxnet

Ran into the following error whilst training a CIFAR model and was hoping that it could be checked before I run the training process again

Traceback (most recent call last):
File "C:\Users\andre\source\repos\3pxnet\3pxnet-training\main_conv.py", line 289, in
main()
File "C:\Users\andre\source\repos\3pxnet\3pxnet-training\main_conv.py", line 269, in main
re = net(testdata)
File "C:\Users\andre\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\andre\source\repos\3pxnet\3pxnet-training\network.py", line 423, in forward
x = self.conv1(x)
File "C:\Users\andre\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\andre\source\repos\3pxnet\3pxnet-training\binarized_modules.py", line 137, in forward
out = nn.functional.conv2d(input, self.weight, None, self.stride,
RuntimeError: Given groups=1, weight of size [128, 3, 3, 3], expected input[100, 32, 5, 34] to have 3 channels, but got 32 channels instead
make: *** [Makefile:13: train] Error 1