lmeyerov

For posterity commas + explicit args for `docker-compose.yml`: ```yml dask-cuda-worker0:

Thanks @pentschev , that's wonderful to hear wrt CPU core / NUMA node. Exploring that was on my list for today/tmw! This is likely more of a `dask_cudf` q, but...

Yeah, NUMA/locality-aware scheduling is def an R&D-level problem in practice, so I get it :) It'd be cool to be able to setup/automate NUMA tags for a multi-gpu node as...

Just hit the same general issue here: https://github.com/rapidsai/dask-cuda/issues/481 I'm unclear of a generic docker recommended path, so if this even matters

Seeing the same ~3X speedup with seccomp:unconfined..

I'd be happy to try upgrading to confirm the fix. Any guidance on how? (apt-get? fresh docker install at some version? basic googling did not make this obvious)

Sorry, haven't tried recently.

Looking more closely at the trace, it looks `FROM rapids...cu11.3` (largely mamba-based) + `pip install torch 1.12.1` + `pip install pyg-lib ...` + `mamba install pyg` was the issue. The...

Financial data gets big quite quickly (millions/billions of transactions, time points, people, ...), so I'm pretty confused why only one format relevant to nation-scale data is considered.

(Most of those formats seem more relevant for messaging, and poor for sharing anything sizeable or allowing compute over it.)