my-spack-envs

My reproducible Spack environments of some HPC platforms I use.

Computing facility	Environment	Spack environments	Description
alcf	nanotron	a100.polaris.spack	Distributed training framework for large language models
alcf	neomem	a100.polaris.spack, a100.thetagpu.spack	Torch rehearsal backend to mitigate catastrophic forgetting with a focus on performance, written in C++
genci	inference	a100.jeanzay.spack	Inference serving and deployment stack based on vLLM
genci	skao		Square Kilometre Array Observatory data processing pipeline
grid5000	neomem	p100.chifflot.spack, v100.chifflot.spack, v100.gemini.spack	Torch rehearsal backend to mitigate catastrophic forgetting with a focus on performance, written in C++
generic	ai-inference		AI inference stack with packages like vLLM
generic	gysela-mini-app-io		A minimal application demonstrating GYSELA operations
generic	kénotron		Experimental fork of Nanotron, a minimalistic large language model 4D-parallelism training
generic	neomem		Torch rehearsal backend to mitigate catastrophic forgetting with a focus on performance, written in C++
generic	pdi-tutorial		Tutorial environment for the PDI data interface library
generic	pdidev		PDI development environment and spack-based CI test images (16 variants)

Workflows

I like to work on this repository from inside a container:

podman run -v /path/to/my-spack-envs:/root/my-spack-envs -v /path/to/spack-packages:/root/.spack/package_repos/fncqgg4 -it docker.io/spack/rockylinux9:develop

You can also use a more stable Spack version:

podman run -v /path/to/my-spack-envs:/root/my-spack-envs -v /path/to/spack-packages:/root/.spack/package_repos/fncqgg4 -it docker.io/spack/rockylinux9

To update package recipes, simply use spack repo update.