ModuleNotFoundError: No module named 'xfuse' #70
Comments
|
Hi Nicolaas, The error message seems to indicate that xfuse wasn't found in your python environment. It could be that something failed during the installation and that xfuse didn't get installed properly or that there is some kind of environment conflict. Can you try running the installation command again and post the output here? If it completes without errors, what is the output of |
|
To install I run
Errors thrown in bold - Really think its a problem with the way I don't know how to handle github installations on the supercomputer. I also asked the supercomputer support team for help - I will keep you updated from my side, but if you know what is the problem, please let me know Nicolaas |
|
Thanks for the update Nicolaas, do let us know what you hear from the support team. I haven't used the |
|
I ran and got this answer:
when running It throws the same error:
|
|
Great! I think those warnings are fine, at least it seems the import is working with the new PYTHONPATH. This modifies the environment for the prefixed command only AFAIK. Can you try prefixing the |
|
I run It says:
I don't understand what you mean by prefixing, but I assumed you meant this: Then he goes:
` R is a bit tricky in the beginning but then its just childsplay. Give me these linux systems and I'm caught in never-ending shitstorm of errors... This is so frustrating... I just don't understand the underlying data structure |
|
Hah, yes I feel you, package and dependency management in Linux can indeed be frustrating at times :) That first error suggests that your shell can no longer find the xfuse executable, maybe you are running the command from a different directory? Alternatively you can always write out the full path: That second command looks good but I think you are missing a space between the environment modifier and the command: My understanding here is that Python is reading the environment variable |
|
calling well... maybe not 100% but we are getting there. Now he throws the following error: /
What does he mean by 'no such option: --counts'? Am I doing sth wrong in the syntax? |
|
Great, looks promising :) xfuse convert has different subcommands for the type of data you are converting ( |
|
omfg its running *crying emoji |
|
Right. Now the next challenge; getting the GPU operational ;) Thanks for the help thusfar! Much appreciated!!! Quick questin; If you stop and re-run the analysis, you seem to start a new epoch. Does that affect the quality in any way? And if I want to run the analysis in a shorter time-frame (quick and dirty, just to see some output fast instead of running for multiple days), would reducing the epoch number from 100k to 10k do the trick? Or would it be better to reduce the scale to .01 during pre-processing? |
I'd guess this depends on the HPC environment (if you are using a workload manager etc.). For the Python installation, just make sure that you have a torch version with cuda support (visible as "+cuXX" in the version string): You can check if the GPU is available like so:
Do you mean starting the run anew or when restoring a previous run (https://github.com/ludvb/xfuse#stopping-and-resuming-a-run)? In general, I wouldn't be concerned so much about starting a new epoch, since the data is shuffled/sampled randomly. More important will be the number of steps taken by the optimizer.
This is a good question. I haven't really found a good way to speed up the analysis frankly. Reducing the scale would indeed be a possibility but could impact the results quite negatively (you will still get some outputs to experiment with, though). What I would suggest is to monitor the training progress using tensorboard (https://github.com/ludvb/xfuse#tracking-the-training-progress), which will give you a good indication of when the model is starting to learn interesting structure. |
I am trying to run xfuse on the supercomputer system; I'm not an expert but I do manage to run other programs.
When I run xfuse
xfuse convert st --counts section1.tsv --image section1.jpg --transformation-matrix section1-alignment.txt --scale 0.15 --save-path section1I get the following:
So the problem here is he thinks
__main__is located in mydir/xfuse/bin/xfuse, while in reality, it is located in mydir/xfuse/The export path is set to the bin directory, otherwise he doesnt find xfuse. But now it seems like the other files cannot be found
What am I doing wrong here???
Kind regards,
Nicolaas
The text was updated successfully, but these errors were encountered: