-
Notifications
You must be signed in to change notification settings - Fork 482
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
won't start #33
Comments
As an update, I got things this far, but am still stuck (I am ATI/AMD so NumPy is fine, I don't care about speed, I just want to get this working): CUDArray: Failed to load CUDA back-end. What am I doing wrong? Definitely something! |
Can you try to make sure the backend is set to Numpy and not Cuda. The error message says that it's trying to load the Cuda libraries |
Thanks! I'm not quite sure how to do that, but if I unset CUDARRAY_BACKEND, my test fails completely - windows error, no console output at all. Same thing if i set it to numpy. If I set it to cuda, I get the error message from my comment above (ArrayData import fail). |
The result of "import cudarray; print(cudarray._backend)" is "numpy". |
Did you run that command to install Cudarray? |
I'm running it again! |
(Because I've tried the CUDA install a few different times now.) Anyway, that setup did not create any setup error messages. But I still get an error when I try to generate an image. And I'm possibly making things worse by trying to fix this in various ways and not remembering which things I might have undone. I think I've spent 4 hrs. Where I'm at now: verify-backend.py: output: |
What strikes me as interesting about my situation:
How weird that 1= "no cuda" and 2="cuda". This does not seem right. But it's the only combination that's even given me output to even get the error message about ArrayData. And for the life of me, I can't google or find anything about ArrayDatay in conda.exe or anywhere. |
I've faced a lot of installation issues on Ubuntu (I'm a big Windows guy...) but in the end I made it work. It took me a while too. |
I have a hard time with things like these, even though I'm a command-line I think this might have been the error: Faulting application name: python.exe, version: 0.0.0.0, time stamp: and when I reproduced it, this is what popped up: Problem signature: Thanks :) |
but anyway, the error that i'd actually like to try to fix is this one: ImportError: No module named array_data ...Because that's the situation where DeepStyle is the closest to running -- it actually produces STDOUT at least :) |
I'd let you remote into my machine, too, if it would help. That might be a nightmare though. It seems like the error is simply that some module called array_data is not being loaded. It's weird because I can't find reference to array_data anywhere. I really wanted to start running this on permutated sets of images... |
It just won't start on windows. Perhaps because I don't know where to put the "cuda" folder, since the installation doesn't say how to add that part to the rest?
S:\new\DeepStyle\neural_artistic_style-master>neural_artistic_style.py --subject image.jpg --style s
tyle.jpg --output output.jpg
CUDArray: CUDA back-end not available, using NumPy.
At that point, I get a windows error:
Problem signature:
Problem Event Name: APPCRASH
Application Name: python.exe
Application Version: 0.0.0.0
Application Timestamp: 56abcaee
Fault Module Name: ntdll.dll
Fault Module Version: 6.1.7601.19160
Fault Module Timestamp: 56bcd74c
Exception Code: c0000005
Exception Offset: 000000000004a604
OS Version: 6.1.7601.2.1.0.256.1
Locale ID: 1033
Additional Information 1: 3935
Additional Information 2: 39350178b1f7dc4d15d5471cf2a5a09f
Additional Information 3: 6976
Additional Information 4: 697614677dd37f475b8d30bce4aa474e
Seems like perhaps an exception is nto being caught?
The text was updated successfully, but these errors were encountered: