I've been trying to install brainiak and use it in the Jupyter notebook. It seems to have installed fine via conda but when I try to import it, it can't find the module
Do you have any suggestions as to how to fix this issue?
I've also tried to install brainiak with pip
but it also didnot work
Did you install brainiak in a conda environment that’s not your root env? Last time I checked, Jupyter doesn’t play nicely with conda envs by default. You need https://docs.anaconda.com/anaconda/user-guide/tasks/use-jupyter-notebook-extensions/#notebook-conda or another similar workaround.
@mihaic and I entered my token here: http://127.0.0.1:8899/login?next=%2Fnotebooks%2Fexamples%2Ffuncalign%2Frsrm_synthetic_reconstruction.ipynb
@mihaic Actually, I just installed condo and followed this link to install brainiak as well: https://brainiak.org
And it works but when I tried the import functions, it said the command cannot be found. I saw that there are other people struggling with that too and another person has answered it by offering this website: https://docs.anaconda.com/anaconda/user-guide/tasks/use-jupyter-notebook-extensions/#notebook-conda
I tried it but again, I couldn't use the function "jupyter"... Would you be kindly to give me some guidance on that? Thanks.
And it works but when I tried the import functions, it said the command cannot be found. I saw that there are other people struggling with that too and another person has answered it by offering this website: https://docs.anaconda.com/anaconda/user-guide/tasks/use-jupyter-notebook-extensions/#notebook-conda
I tried it but again, I couldn't use the function "jupyter"... Would you be kindly to give me some guidance on that? Thanks.
@mihaic Okay, so here's my progress with Docker and please bear with me because I'm relatively new to this... So I ran "docker pull brainiak/brainiak" and it went without error, but when I ran "ocker run -it -p 8899:8899 --name demo brainiak/brainiak", it showed this:
The container name "/demo" is already in use by container "db4b66acb135f4832b75dea2fcc1d67abdca636a6cd1cb9a34801de6066aa49e". You have to remove (or rename) that container to be able to reuse that name.
The container name "/demo" is already in use by container "db4b66acb135f4832b75dea2fcc1d67abdca636a6cd1cb9a34801de6066aa49e". You have to remove (or rename) that container to be able to reuse that name.
So I think that hyphenated random codes are the token so I pasted that onto the Jupiter notebook page but it did not work...
Then the run command should work again and you should see a new URL.
Alternatively, you can force remove the container:
docker container rm -f demo
Hello! I was working on Brainiak tutorial #10 and was wondering what I could do to get the answer key? There are lot of sections left out and I would love to be able to see them!
@jane2qian I don't think SRM has GPU support. Parallelization is achieved by MPI. Other people can correct me if I'm wrong about it. Though I think we are about to have a faster version of SRM: brainiak/brainiak#416 ; btw how many voxels/TRs do you have in your data per participant?
@sncraig01 At present we have no answer key. We are making some edits that will be finalized soon. After that we will give the answer key to anyone who promises that they are not currently nor in the future will be in a for-credit course that uses these materials. In the mean time, if you would like me to share any of the answers with you then you can reach out to me directly and ask any specific questions you have
Thank you! @qihongl 166953 voxels*973 TRs per subject. Our aim is to differentiate groups using two step srm. I wanna try several parameters to get the optimized model in each step. I should just use the old srm version then until the fast version is released. Any advice on the parameters selection?
Hi, @jane2qian sorry about the delayed reply... I forgot to check this thread.
I think the number of voxels suggests you are doing some whole-brain analysis? I don't think the amount of data you have is enough for SRM to estimate Ws and S. Is it possible to constrain the analysis to certain ROIs? This is what's typically done in the literature.
I think the number of voxels suggests you are doing some whole-brain analysis? I don't think the amount of data you have is enough for SRM to estimate Ws and S. Is it possible to constrain the analysis to certain ROIs? This is what's typically done in the literature.
I am having an issue installing Brainiak. I am doing it on MacOS, and even tried using a new environment (as the person above who last had trouble). I keep getting to the failure to build a wheel, but it then always fails to install after: "ERROR: Failed building wheel for brainiak
Running setup.py clean for brainiak
Failed to build brainiak
ERROR: Could not build wheels for brainiak which use PEP 517 and cannot be installed directly"
Running setup.py clean for brainiak
Failed to build brainiak
ERROR: Could not build wheels for brainiak which use PEP 517 and cannot be installed directly"
when I try it via Conda "conda install -c brainiak -c defaults -c conda-forge brainiak" It keeps doing "Collecting package metadata (current_repodata.json): done
Solving environment: failed with current_repodata.json, will retry with next repodata source.
Initial quick solve with frozen env failed. Unfreezing env and trying again." and then doesn't seem to ever succeed
Solving environment: failed with current_repodata.json, will retry with next repodata source.
Initial quick solve with frozen env failed. Unfreezing env and trying again." and then doesn't seem to ever succeed
Hello, @zacharybretton. Thanks for the feedback. Regarding the PEP 517 error, try adding the --no-use-pep517 to the install command:
python3 -m pip install --no-use-pep517 brainiakSee for more details brainiak/brainiak#435
We'll look into the problem with the newest Conda version.
HI Mingbo. The searchlight code natively deals with returning a list inside the kernel function. This means the output will be a numpy array with each voxel being a list. However because of the nature of numpy arrays this can be difficult to index. If you want some example code of how to use this output then I am happy to send it to you.
Sorry about the delay @lcnature, I got buried at SFN. Here is a small chunk of code you can use that should help:
masked_data = sl_result[mask == 1] # Get only the masked data
number_elements = len(masked_data[0]) # Get the length of the first unmasked voxel
concat_sl_result = np.zeros((sl_result.shape) + (number_elements, )) # Concatenate tuples to get a stacked volume for each of the SL kernel outputs
# Cycle through the outputs of the SL kernel
for element_counter in range(number_elements):
# Get the volume for this SL kernel
concat_sl_result[mask == 1, element_counter] = [voxel[element_counter] for voxel in masked_data]