env creation fails

[RobertoDF]

Hi,

Creating a conda env using mamba env create -f environment.yml fails with:

Could not solve for environment specs
Encountered problems while solving:
  - package python-3.6.8-h0371630_0 requires libffi >=3.2.1,<3.3a0, but none of the providers can be installed
  - package python-3.6.7-h0371630_0 requires libffi >=3.2.1,<3.3a0, but none of the providers can be installed

[alexpiet]

Hi @RobertoDF, sorry for the trouble!

In general, creating that environment is going to be difficult because a lot of dependencies are pinned so I can recreate the paper figures. Depending on what you want to do with the code, you probably dont need most of the packages, and certainly dont need specific pinned versions of most of them. I'd recommend making your own environment.yml with the bare-bones needed.

Note that if you want to analyze the model fits I already made, you only need pandas to load the files. Happy to help more if you have specific questions

[RobertoDF]

Hi Alex, Thanks for answering! I am interested in running this analysis on the Neuropixel dataset. I also have a collab with the Allen Institute via Openscope and I would be interested in tracking your strategy index across sessions (we´ll also use the change detection task). I see that there is a ad hoc repo for that (https://github.com/AllenInstitute/licking_behavior_NP). Are these parameters the one you used in your final analysis?

format_options = {
                'timing0/1':True,
                'mean_center':True,
                'timing_params':[-5,4],
                'timing_params_session':[-5,4],
                'ignore_trial_errors':False,
                'num_cv_folds':10,
                'git_commit_hash': git_hash,
                'git_branch':git_branch
                }

I get a error during the last cross validation round. Do you have any idea why is that? Did somebody already processed that dataset using your model?
Thanks, Roberto

[alexpiet]

Great to hear you are interested in using this code on the NP dataset and an Openscope task. You should use the licking_behavior_NP repo, since there is some NP specific processing (mostly in how spikes get binned into timesteps). Yes, I already ran the model on the NP dataset. If you give me your email (or email me alexpiet at gmail), then I can send you the figures, and the fit models.

[alexpiet]

format_options = {
               'timing0/1':True,
               'mean_center':True,
               'timing_params':[-5,4],
               'timing_params_session':[-5,4],
               'ignore_trial_errors':False,
               'num_cv_folds':10,
               'git_commit_hash': git_hash,
               'git_branch':git_branch
               }

When running the NP dataset I used:

format_options = {
               'timing0/1':True,
               'mean_center':True,
               'timing_params':[-5,4],
               'timing_params_session':[-5,4],
               'ignore_trial_errors':False,
               'num_cv_folds':10,
               'git_commit_hash': ddcfd4b,
               'git_branch':master
               }