[Ubuntu14.04,python2.7.5,CUDA6.5] invalid device function..cudamat.cudamat.CUDAMatException: CUDA error: no error

[lireagan]

I got this problem, And I think it is not about CUDA, because I had make "NVIDIA_CUDA-6.5_Samples", and run the deviceQuery. The result is pass. I don't know why this happens and it has trouble me for weeks. PLEASE help me ! THX A LOT!

[lireagan]

I hava figure it out.. Just edit the Makefile

[interxuxing]

Really helps! Thank you for your sharing. If someone use cuda > 5.5, this modification of the Makefile is necessary.

[IraitzM]

Thanks for sharing! Actually I have quite the same problem but not exactly the same :) for a GTX970 I added -gencode arch=compute_52,code=sm_52 and I am still getting same error at line 640 (add_sums function) but not always :S. Every once in a while it works!

If anybody has any idea how this issue could be fixed, it would be more than appreciated :)

[YantianZha]

But I still had error, after "make" that:

My Makefile:

Thanks in advance!

[guxiaoling]

@YantianZha, you may forget to add the following lines to your ~/.bashrc file (notice your cuda version) export CUDA_BIN=/usr/local/cuda-7.5/bin
export CUDA_LIB=/usr/local/cuda-7.5/lib64

[buaa-luzhi]

@IraitzM hello, Did you solve the problem, My cuda version is 8.0, This is my question. Traceback (most recent call last): File "lstm_combo.py", line 405, in cm.CUDAMatrix.init_random(42) File "/home/e829/Documents/LuZhi/20161213/unsupervised-video/cudamat/cudamat.py", line 382, in init_random raise generate_exception(err_code) cudamat.cudamat.CUDAMatException: CUDA error: no error

Any one can help me! Thank you very much!

[lizhijianrd]

@buaa-luzhi Have you solved the problem? I got the similar problem with you when running ae, here are the logs: Autoencoder 1 invalid device function Traceback (most recent call last): File "../../trainer.py", line 60, in main() File "../../trainer.py", line 53, in main model = CreateDeepnet(model, train_op, eval_op) File "../../trainer.py", line 36, in CreateDeepnet return NeuralNet(model, train_op, eval_op) File "/home/snail/bash_workspace/1202/deepnet/deepnet/neuralnet.py", line 43, in init cm.CUDAMatrix.init_random(self.net.seed) File "/home/snail/bash_workspace/1202/deepnet/cudamat/cudamat.py", line 224, in init_random raise generate_exception(err_code) cudamat.cudamat.CUDAMatException: CUDA error: no error

My cuda version is 8.0, do you know how to solve this problem?

Thank you very much!