search

Hanjun-Dai / graph_comb_opt

Implementation of "Learning Combinatorial Optimization Algorithms over Graphs"
https://arxiv.org/abs/1704.01665
MIT License
491 stars 135 forks source link

New clone fails due to lack of graphnn permission #12

Open Divide-By-0 opened 5 years ago

Divide-By-0 commented 5 years ago

When I run the first command, [..]> git clone --recursive https://github.com/Hanjun-Dai/graph_comb_opt I get this error, which seems to be related to outdated ssh keys online?

Cloning into 'graph_comb_opt'...
remote: Enumerating objects: 6, done.
remote: Counting objects: 100% (6/6), done.
remote: Compressing objects: 100% (5/5), done.
remote: Total 271 (delta 1), reused 3 (delta 1), pack-reused 265
Receiving objects: 100% (271/271), 2.28 MiB | 2.52 MiB/s, done.
Resolving deltas: 100% (134/134), done.
Submodule 'graphnn' (git@github.com:Hanjun-Dai/graphnn) registered for path 'graphnn'
Cloning into '[..]/graph_comb_opt/graphnn'...
git@github.com: Permission denied (publickey).
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.
fatal: clone of 'git@github.com:Hanjun-Dai/graphnn' into submodule path '[..]/graph_comb_opt/graphnn' failed
Failed to clone 'graphnn'. Retry scheduled
Cloning into '[..]/graph_comb_opt/graphnn'...
git@github.com: Permission denied (publickey).
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.
fatal: clone of 'git@github.com:Hanjun-Dai/graphnn' into submodule path '[..]/graph_comb_opt/graphnn' failed
Failed to clone 'graphnn' a second time, aborting
Hanjun-Dai commented 5 years ago

I checked on my side and it works well.

Probably that is relevant to your own key?

You can also clone the two repositories separately, without using recursive clone.

Divide-By-0 commented 5 years ago

Sure, so if I just drag the other repo into this graphnn and clone non recursively, it should be the same right?

Hanjun-Dai commented 5 years ago

It should be. But probably you want to make sure the version of graphnn matches (graphnn @ bdf51e6)

Divide-By-0 commented 5 years ago

Sure, the only change since then is cuda being restricted to 8.0. Now I get the error when running make -j on realworld tsp2d: make: *** No rule to make target ../../../graphnn/build/lib/libgnn.a', needed bybuild/dll/libtsp2d.so'. Stop.

Hanjun-Dai commented 5 years ago

did you 1) place graphnn in the right folder? 2) get graphnn built successfully first?

Divide-By-0 commented 5 years ago

So I'm trying to build graphnn and theres this error since I don't have a GPU. I'll switch to a machine with one then keep you updated.

In file included from include/tensor/cpu_dense_tensor.h:5:0, from include/tensor/tensor_all.h:5, from include/nn/variable.h:7, from include/nn/factor.h:4, from include/nn/cross_entropy.h:5, from src/nn/cross_entropy.cpp:1: include/tensor/t_data.h:6:34: fatal error: thrust/device_vector.h: No such file or directory

include <thrust/device_vector.h>

                              ^
daltunay commented 2 years ago

Hi, I have the same error. I successfully built graphnn, and placed it manually (due to no permission error) in the graph_comb_opt folder. When I run make -j, I get the error : make: *** No rule to make target '../../../graphnn/build/lib/libgnn.a', needed by 'build/dll/libmvc.so'. Stop.

I don't know if that's important, but /graphnn/build only contains an objs folder, and no dll one. At least I can't see it. Though, I was able to run the test on MNIST with graphnn.

Any ideas ?