Issues with mesmerize Execution: Missing Packages in Active Conda Environment

[Tepps-ps]

After installing mesmerize and its dependencies in a Conda environment, I encounter multiple warnings and a ModuleNotFoundError when attempting to launch mesmerize. Despite successful installation (verified through conda list), mesmerize reports that it cannot find caiman, tensorflow, tslearn, and h5py not being found. Any advice on further steps or potential solutions would be greatly appreciated.

Environment: OS: Windows 10 Conda version: 23.11.0 Python version: 3.10.13 Installed packages: caiman 1.10.2 tensorflow 2.10.0 tslearn 0.6.3 h5py 3.9.0

[kushalkolar]

What installation instructions did you follow?

[Tepps-ps]

I followed the installation instructions with the following steps:

conda install -c conda-forge mamba mamba create -n mescore -c conda-forge mesmerize-core mamba activate mescore caimanmanager install pip install "fastplotlib[notebook]"

[kushalkolar]

If you activate the environment and launch jupyterlab do the demo notebooks run?

[kushalkolar]

I notice you deleted a comment about the kernel dieing (we get notifications about everything, you don't have to be shy :P ). Either you have a GPU with insufficient RAM or you don't have a modern GPU.

[Tepps-ps]

Thank you for finding the comment! Updating the driver resolved the issue with the kernel crashing. However, when I executed the command 'df.iloc[0].caiman.run(),' I encountered an error stating, 'ValueError: need at most 63 handles, got a sequence of length 65.' The process has been running for over a dozen hours. I'm using Windows; doesn't this seem like an excessively long time?

[kushalkolar]

I need to see the full traceback

On Fri, Feb 16, 2024, 03:09 Tepps-ps @.***> wrote:

Thank you for finding the comment! Updating the driver resolved the issue with the kernel crashing. However, when I executed the command 'df.iloc[0].caiman.run(),' I encountered an error stating, 'ValueError: need at most 63 handles, got a sequence of length 65.' The process has been running for over a dozen hours. I'm using Windows; doesn't this seem like an excessively long time?

— Reply to this email directly, view it on GitHub https://github.com/nel-lab/mesmerize-core/issues/272#issuecomment-1947926957, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACHXXRA2UDK6B6HQFCIUJC3YT4HZ5AVCNFSM6AAAAABC7ILMYWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNBXHEZDMOJVG4 . You are receiving this because you commented.Message ID: @.***>

[Tepps-ps]

Here is the traceback. It has been more than 12 hours since running the command.

[kushalkolar]

it says motion correction finished successfully?

[Tepps-ps]

Yes, but still running..

[kushalkolar]

Send a keyboard interrupt

[Tepps-ps]

After keyboard interrupt, an error log about multiprocessing was displayed.

[kushalkolar]

yea that happens when you do a keyboard interrupt, are you able to view the results? What are the dims of the input movie?

[Tepps-ps]

I'm unable to view any results. I used caimadata examples named 'data_endoscope'. It consists of 1000 frames, 128x128 pixels and a 16-bit depth.

[kushalkolar]

I'm unable to view any results.

Can you please elaborate, what did you do and what do you get?

[Tepps-ps]

Basically, the commands prepared in advance in the demo notebook 'cnmfe.ipynb' are executed in order from the top. While running the CaIman run command (df.iloc[0].caiman.run()) for motion correction on the first item in the DataFrame, the Jupyter Notebook remained in the "running" state for over 12 hours, even though the log output indicated that the process had finished successfully with the message mc finished successfully!. The DataFrame's first item (index 0) was configured with the following parameters:

• Algorithm: mcorr (motion correction)
• Item name: data_endoscope
• Input movie path: example_movies\data_endoscope.tif
• Parameters: {'gSig_filt': (3, 3), 'pw_rigid': True, 'max_shifts': (5, 5), 'strides': (48, 48), 'overlaps': (24, 24),,,

In the output directory, files associated with the UUID d30a6e53-f727-4fee-97eb-c69b1ab8a027 were generated, which suggests that some level of processing was completed. However, a ValueError indicating that more than 63 handles were being waited on was encountered, which may be related to the hangup.

[kushalkolar]

Exactly what files are present in the uuid dir within the batch dir? How did you try to view the output? Please provide screenshots and as much information as you can. I unfortunately cannot help without all the details :)

[Tepps-ps]

Sorry for bothering you. Here is screenshot of files are present in the uuid dir. And I also shared the results of running the command df = df.caiman.reload_from_disk() in the Jupytert lab. I thought that if df.iloc[0].caiman.run() completed successfully, some value would be added in index[0]output is this correct?

[kushalkolar]

From this it seems like multiprocessing on Windows doesn't work with more than 60 processes: https://github.com/scikit-learn/scikit-learn/issues/13354

You can limit the number of processes spawned to less than 60, see this: https://github.com/nel-lab/mesmerize-core/issues/205#issuecomment-1574585472

[Tepps-ps]

Thank you for your prompt and courteous response (I apologize for the delayed reply). The issue was resolved by limiting the number of processes spawned to less than 60.