Error on running Bambi on Colab

Greetings all,

I am trying to run bambi on colab, however I got into some issues;

First problem is as follows;

`NoSectionError                            Traceback (most recent call last)
[/usr/local/lib/python3.7/dist-packages/aesara/configparser.py](https://localhost:8080/#) in fetch_val_for_key(self, key, delete_key)
    236             try:
--> 237                 return self._aesara_cfg.get(section, option)
    238             except InterpolationError:

15 frames
NoSectionError: No section: 'blas'

During handling of the above exception, another exception occurred:

KeyError                                  Traceback (most recent call last)
KeyError: 'blas__ldflags'

During handling of the above exception, another exception occurred:

ModuleNotFoundError                       Traceback (most recent call last)
ModuleNotFoundError: No module named 'mkl'

During handling of the above exception, another exception occurred:

RuntimeError                              Traceback (most recent call last)
[/usr/local/lib/python3.7/dist-packages/aesara/link/c/cmodule.py](https://localhost:8080/#) in check_mkl_openmp()
   2683 you set this flag and don't set the appropriate environment or make
   2684 sure you have the right version you *will* get wrong results.
-> 2685 """
   2686         )
   2687 

RuntimeError: 
Could not import 'mkl'.  If you are using conda, update the numpy
packages to the latest build otherwise, set MKL_THREADING_LAYER=GNU in
your environment for MKL 2018.

If you have MKL 2017 install and are not in a conda environment you
can set the Aesara flag blas__check_openmp to False.  Be warned that if
you set this flag and don't set the appropriate environment or make
sure you have the right version you *will* get wrong results.`

Than I’ve updated mkl via !pip install mkl -U and got the second issue;

AttributeError                            Traceback (most recent call last)
[<ipython-input-7-e7c20f48c217>](https://localhost:8080/#) in <module>()
----> 1 import bambi as bmb

6 frames
[/usr/local/lib/python3.7/dist-packages/aesara/tensor/nnet/opt.py](https://localhost:8080/#) in <module>()
    492 
    493 # Register Cpu Optimization
--> 494 conv_groupopt = aesara.graph.optdb.LocalGroupDB()
    495 conv_groupopt.__name__ = "conv_opts"
    496 register_specialize_device(conv_groupopt, "fast_compile", "fast_run")

AttributeError: module 'aesara' has no attribute 'graph'

Then, I’ve update tensorflow with !pip install -U tensorflow-gpu and I got stuck in third;

ValueError                                Traceback (most recent call last)
[<ipython-input-9-e7c20f48c217>](https://localhost:8080/#) in <module>()
----> 1 import bambi as bmb

8 frames
[/usr/local/lib/python3.7/dist-packages/aesara/graph/optdb.py](https://localhost:8080/#) in register(self, name, optimizer, use_db_name_as_tag, *tags, **kwargs)
     69 
     70         if name in self.__db__:
---> 71             raise ValueError(f"The tag '{name}' is already present in the database.")
     72 
     73         if use_db_name_as_tag:

ValueError: The tag 'local_inplace_sparse_block_gemv' is already present in the database.

Anybody has the same problem? I am not expert enough to proceed further, can somebody explain what is actually wrong?

Thanks.

[canyon289]

To scope this ticket, it sounds like the error for running Bambi on colab has been resolved right? We have a short term fix, and I assure you the long term fix is actively being worked on. If the issue now is slow sampling on colab, thats a very different thing, either requiring model debugging, data thinning, or frankly just paying for colab pro to get more computational speed.

Let me know if my assessment here is correct

[hans-ekbrand]

Now I am trying to run sampling on google colab (standart) with GPU

warmup: 6%|▋ | 699/11000 [37:59<**9:42:17**, 3.39s/it, 1023 steps of size 2.09e-03. acc. prob=0.93]

I am not sure how I should proceed, and I am open to suggestions.

Collect samples semi-manually: set “draws” to a number low enough so that the run finish before the session is interrupted, save the arviz object to disk (with pickle), transfer the file to a persistant storage medium (colab is wiped automatically when you get disconnected), disconnect the colab instance, wait until your quota on colab is high enough for you to aquire a new GPU, repeat. Do posterior analys only when you have collected enough draws.

EDIT: no reason to use “parallel” on colab, you only have one GPU there.