dask-memusage

dask-memusage copied to clipboard

Typo in front page doc CLI usage line.

1

Change ``` $ dask-scheduler --preload dask_memusage --memusage.csv /tmp/memusage.csv ``` to ``` $ dask-scheduler --preload dask_memusage --memusage-csv /tmp/memusage.csv ```

y-he2

Add dask_memusage.install introduce "ValueError: Inputs contain futures that were created by another client."

2

Thank you for the wonderful tool! I would like to profile peak memory of my dask application. I can run it successfully without dask_memusage. However, after I add memusage.install, it...

CAROLZXYZXY

Plus other tweaks. Fixes #7.

itamarst

itamarst

Before graph is run, optimizer may connect tasks. Profiling is per optimized task.

itamarst

If you want 8 worker processes: ```python from dask.distributed import Client, LocalCluster cluster = LocalCluster(n_workers=8, threads_per_worker=1, memory_limit=None) Client(cluster) ```

itamarst

* [x] Final cleanup of README * [x] Do the release * [ ] Announce it (Dask list? Dask blog? Elsewhere?)

itamarst

If it's possible to detect this, we should exit loudly if there are more than 1 thread per worker process, since that invalidates the results.

itamarst

Today, it is almost impossible to monitor GPU usage using dask_memusage. Only regular memory is possible. With the proposed commits it is possible at least to compare the minimum GPU...

jcfaracco