setup.py and .pth

[yunshangyue71]

I find there is no setup.py and the pre_trained model which you offer is not match the network.
I think you trained the model on another script which you not offer ,recently you improved your script, but you didn't retrained .

[GoGoDuck912]

1. The current code base adopts the implementation of InplaceSyncBN from https://github.com/mapillary/inplace_abn. The compiling of the InplaceSyncBN operator will be automatically triggered when running the code.

2. Our former codebase is based on pytorch version 0.4. However, we only made minor modifications regarding pytorch version difference, and upgrade to pytorch>1.0 for simplicity usage. From my side, the current provided pre-trained model matches the network configuration. As we have organized the code for easy reading, please do not mix the former code base with the current one. Please clarify what kind of mismatch you met.

[yunshangyue71]

After killing several bugs that produced by my environment.The major bug apper.
(pytorch15) q30004136:/SCHP_org$ python train.py
Traceback (most recent call last):
File "train.py", line 25, in
import networks
File "/data_ssd3/q30004136/SCHP_org/networks/init.py", line 3, in
from networks.AugmentCE2P import resnet101
File "/data_ssd3/q30004136/SCHP_org/networks/AugmentCE2P.py", line 21, in
from modules import InPlaceABNSync
File "/data_ssd3/q30004136/SCHP_org/modules/init.py", line 1, in
from .bn import ABN, InPlaceABN, InPlaceABNSync
File "/data_ssd3/q30004136/SCHP_org/modules/bn.py", line 10, in
from .functions import *
File "/data_ssd3/q30004136/SCHP_org/modules/functions.py", line 18, in
extra_cuda_cflags=["--expt-extended-lambda"])
File "/data_ssd3/q30004136/anaconda3/envs/pytorch15/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 921, in load
is_python_module)
File "/data_ssd3/q30004136/anaconda3/envs/pytorch15/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 1120, in _jit_compile
return _import_module_from_library(name, build_directory, is_python_module)
File "/data_ssd3/q30004136/anaconda3/envs/pytorch15/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 1441, in _import_module_from_library
file, path, description = imp.find_module(module_name, [path])
File "/data_ssd3/q30004136/anaconda3/envs/pytorch15/lib/python3.7/imp.py", line 296, in find_module
raise ImportError(_ERR_MSG.format(name), name=name)
ImportError: No module named 'inplace_abn'
(pytorch15) q30004136:/SCHP_org$

my environment is CUDA 10.2 ,ubuntu 18.0.4,python3.7
The error is there is no inplace_abn model .Do i need download the inplace_abn from https://github.com/mapillary/inplace_abn
and then put into the project?
when i run this command ：pip install inplace_abn.it says that: Requirement already satisfied: inplace_abn in /data_ssd3/q30004136/anaconda3/envs/pytorch15/lib/python3.7/site-packages/inplace_abn-0.0.0-py3.7-linux-x86_64.egg (0.0.0)

i don't kown how to solve this problem. I need your help please.

[GoGoDuck912]

You don't need to install inplace_abn by yourself. The current code base already contains the inplace_abn implementation in ./modules. And the compiling of the InplaceSyncBN operator will be automatically triggered when running the code.

[yunshangyue71]

I have tried several times ，it always the same error.

File "/data_ssd3/q30004136/SCHP_org/modules/functions.py", line 18, in
extra_cuda_cflags=["--expt-extended-lambda"])
ImportError: No module named 'inplace_abn'

How do i solve this problem？thank you very much！

[amandazw]

@yunshangyue71 ImportError: No module named 'inplace_abn' did you solve this？

[xhxbuaa]

I met the same problem,have you solved it? my environment is cuda11.2 , pytorch 1.5.1
envs/schp/lib/python3.8/imp.py", line 296, in find_module raise ImportError(_ERR_MSG.format(name), name=name) ImportError: No module named 'inplace_abn'