Large Array Handling Error in estimate_place_fields Function

[ncguilbeault]

Description

There is an issue in the estimate_place_fields function in spiking_likelihood_glm.py when using large arrays of spiking data (>2GB). The function fails, and looking through the callstack shows a ValueError: xxx exceeds max_bin_length(2147483647). This error stems from the line results = dask.compute(*results) and is documented in the dask repository several times (for example here and here).

Current Behavior

When using large spike arrays (>2GB), the function fails with the aforementioned ValueError.

Expected Behavior

The function should be able to handle large spike arrays without throwing an error.

Proposed Solution

A simple fix for this would be to call client.scatter on the spikes array before passing it to the dask compute cluster.

Steps to Reproduce

Use the estimate_place_fields function with a spike array larger than 2GB. Observe the ValueError being thrown.

[edeno]

Closed by #37