{2025.06}[foss/2025a] RELION-5.1.0-foss-2025a-CUDA-12.8.0.eb

[jpecar]

Try newer version of Relion. Challenges expected.

[Neves-P]

Hi @jpecar, thanks for the PR! There is a number of CPU only dependencies that we need to get on EESSI first before we build this, and it's better that we work through the big dependencies first (like PyTorch) before we start work here, so I'll mark this as a draft PR for now.

102 out of 318 required modules missing:

* CUDA/12.8.0 (CUDA-12.8.0.eb)
* einops/0.8.1-GCCcore-14.2.0 (einops-0.8.1-GCCcore-14.2.0.eb)
* parameterized/0.9.0-GCCcore-14.2.0 (parameterized-0.9.0-GCCcore-14.2.0.eb)
* utf8proc/2.10.0-GCCcore-14.2.0 (utf8proc-2.10.0-GCCcore-14.2.0.eb)
* flit-core/3.12.0-GCCcore-14.2.0 (flit-core-3.12.0-GCCcore-14.2.0.eb)
* typing-extensions/4.14.0-GCCcore-14.2.0 (typing-extensions-4.14.0-GCCcore-14.2.0.eb)
* pydantic/2.11.7-GCCcore-14.2.0 (pydantic-2.11.7-GCCcore-14.2.0.eb)
* tlparse/0.4.3-GCCcore-14.2.0 (tlparse-0.4.3-GCCcore-14.2.0.eb)
* googlebenchmark/1.9.4-GCCcore-14.2.0 (googlebenchmark-1.9.4-GCCcore-14.2.0.eb)
* pytest-flakefinder/1.1.0-GCCcore-14.2.0 (pytest-flakefinder-1.1.0-GCCcore-14.2.0.eb)
* pytest-rerunfailures/16.1-GCCcore-14.2.0 (pytest-rerunfailures-16.1-GCCcore-14.2.0.eb)
* pytest-shard/0.1.2-GCCcore-14.2.0 (pytest-shard-0.1.2-GCCcore-14.2.0.eb)
* pytest-subtests/0.15.0-GCCcore-14.2.0 (pytest-subtests-0.15.0-GCCcore-14.2.0.eb)
* optree/0.18.0-GCCcore-14.2.0 (optree-0.18.0-GCCcore-14.2.0.eb)
* Abseil/20250512.1-GCCcore-14.2.0 (Abseil-20250512.1-GCCcore-14.2.0.eb)
* RE2/2024-07-02-GCCcore-14.2.0 (RE2-2024-07-02-GCCcore-14.2.0.eb)
* protobuf/31.1-GCCcore-14.2.0 (protobuf-31.1-GCCcore-14.2.0.eb)
* protobuf-python/6.31.1-GCCcore-14.2.0 (protobuf-python-6.31.1-GCCcore-14.2.0.eb)
* grpcio/1.76.0-GCCcore-14.2.0 (grpcio-1.76.0-GCCcore-14.2.0.eb)
* Arrow/22.0.0-gfbf-2025a (Arrow-22.0.0-gfbf-2025a.eb)
* morphosamplers/0.0.16-gfbf-2025a (morphosamplers-0.0.16-gfbf-2025a.eb)
* mrcfile/1.5.4-gfbf-2025a (mrcfile-1.5.4-gfbf-2025a.eb)
* cuDNN/9.10.1.4-CUDA-12.8.0 (cuDNN-9.10.1.4-CUDA-12.8.0.eb)
* absl-py/2.3.1-GCCcore-14.2.0 (absl-py-2.3.1-GCCcore-14.2.0.eb)
* tensorboard/2.20.0-gfbf-2025a (tensorboard-2.20.0-gfbf-2025a.eb)
* unittest-xml-reporting/3.1.0-GCCcore-14.2.0 (unittest-xml-reporting-3.1.0-GCCcore-14.2.0.eb)
* cuSPARSELt/0.8.0.4-CUDA-12.8.0 (cuSPARSELt-0.8.0.4-CUDA-12.8.0.eb)
* PyQt-builder/1.18.1-GCCcore-14.2.0 (PyQt-builder-1.18.1-GCCcore-14.2.0.eb)
* UCX-CUDA/1.18.0-GCCcore-14.2.0-CUDA-12.8.0 (UCX-CUDA-1.18.0-GCCcore-14.2.0-CUDA-12.8.0.eb)
* libclc/20.1.8-GCCcore-14.2.0 (libclc-20.1.8-GCCcore-14.2.0.eb)
* NCCL/2.27.7-GCCcore-14.2.0-CUDA-12.8.0 (NCCL-2.27.7-GCCcore-14.2.0-CUDA-12.8.0.eb)
* Triton/3.5.0-gfbf-2025a-CUDA-12.8.0 (Triton-3.5.0-gfbf-2025a-CUDA-12.8.0.eb)
* tcsh/6.24.16-GCCcore-14.2.0 (tcsh-6.24.16-GCCcore-14.2.0.eb)
* wrapt/1.17.2-GCCcore-14.2.0 (wrapt-1.17.2-GCCcore-14.2.0.eb)
* pyzstd/0.19.0-GCCcore-14.2.0 (pyzstd-0.19.0-GCCcore-14.2.0.eb)
* NVSHMEM/3.3.20-gompi-2025a-CUDA-12.8.0 (NVSHMEM-3.3.20-gompi-2025a-CUDA-12.8.0.eb)
* HMMER/3.4-gompi-2025a (HMMER-3.4-gompi-2025a.eb)
* PyHMMER/0.11.3-gompi-2025a (PyHMMER-0.11.3-gompi-2025a.eb)
* numba/0.62.0-foss-2025a (numba-0.62.0-foss-2025a.eb)
* magma/2.9.0-foss-2025a-CUDA-12.8.0 (magma-2.9.0-foss-2025a-CUDA-12.8.0.eb)
* umap-learn/0.5.9.post2-foss-2025a (umap-learn-0.5.9.post2-foss-2025a.eb)
* bcrypt/4.3.0-GCCcore-14.2.0 (bcrypt-4.3.0-GCCcore-14.2.0.eb)
* OpenSSL/1.1 (OpenSSL-1.1.eb)
* OpenGL/2025.09-GCCcore-14.2.0 (OpenGL-2025.09-GCCcore-14.2.0.eb)
* CMake/3.18.4 (CMake-3.18.4.eb)
* PyOpenGL/3.1.10-GCCcore-14.2.0 (PyOpenGL-3.1.10-GCCcore-14.2.0.eb)
* cxxopts/3.0.0 (cxxopts-3.0.0.eb)
* PBZIP2/1.1.13-GCCcore-14.2.0 (PBZIP2-1.1.13-GCCcore-14.2.0.eb)
* configurable-http-proxy/5.0.1-GCCcore-14.2.0 (configurable-http-proxy-5.0.1-GCCcore-14.2.0.eb)
* Qt5/5.15.16-GCCcore-14.2.0 (Qt5-5.15.16-GCCcore-14.2.0.eb)
* PyQt5/5.15.11-GCCcore-14.2.0 (PyQt5-5.15.11-GCCcore-14.2.0.eb)
* PycURL/7.45.6-GCCcore-14.2.0 (PycURL-7.45.6-GCCcore-14.2.0.eb)
* gmpy2/2.2.0-GCCcore-14.2.0 (gmpy2-2.2.0-GCCcore-14.2.0.eb)
* sympy/1.14.0-gfbf-2025a (sympy-1.14.0-gfbf-2025a.eb)
* CoinUtils/2.11.12-GCC-14.2.0 (CoinUtils-2.11.12-GCC-14.2.0.eb)
* xpdf/4.06-GCCcore-14.2.0 (xpdf-4.06-GCCcore-14.2.0.eb)
* QtPy/2.4.3-GCCcore-14.2.0 (QtPy-2.4.3-GCCcore-14.2.0.eb)
* Osi/0.108.11-GCC-14.2.0 (Osi-0.108.11-GCC-14.2.0.eb)
* aiohttp/3.12.13-GCCcore-14.2.0 (aiohttp-3.12.13-GCCcore-14.2.0.eb)
* Greenlet/3.2.3-GCCcore-14.2.0 (Greenlet-3.2.3-GCCcore-14.2.0.eb)
* jupyter-server-proxy/4.4.0-GCCcore-14.2.0 (jupyter-server-proxy-4.4.0-GCCcore-14.2.0.eb)
* nbclassic/1.3.1-GCCcore-14.2.0 (nbclassic-1.3.1-GCCcore-14.2.0.eb)
* jupyter-resource-usage/1.1.1-GCCcore-14.2.0 (jupyter-resource-usage-1.1.1-GCCcore-14.2.0.eb)
* Clp/1.17.10-foss-2025a (Clp-1.17.10-foss-2025a.eb)
* Cgl/0.60.9-foss-2025a (Cgl-0.60.9-foss-2025a.eb)
* Cbc/2.10.12-foss-2025a (Cbc-2.10.12-foss-2025a.eb)
* psycopg/3.2.9-GCCcore-14.2.0 (psycopg-3.2.9-GCCcore-14.2.0.eb)
* PuLP/3.3.0-foss-2025a (PuLP-3.3.0-foss-2025a.eb)
* SQLAlchemy/2.0.41-GCCcore-14.2.0 (SQLAlchemy-2.0.41-GCCcore-14.2.0.eb)
* PyTorch/2.9.1-foss-2025a-CUDA-12.8.0 (PyTorch-2.9.1-foss-2025a-CUDA-12.8.0.eb)
* torchvision/0.24.0-foss-2025a-CUDA-12.8.0 (torchvision-0.24.0-foss-2025a-CUDA-12.8.0.eb)
* Faiss/1.13.2-foss-2025a-CUDA-12.8.0 (Faiss-1.13.2-foss-2025a-CUDA-12.8.0.eb)
* PyTorch/2.9.1-foss-2025a (PyTorch-2.9.1-foss-2025a.eb)
* Faiss/1.14.3-foss-2025a-CUDA-12.8.0 (Faiss-1.14.3-foss-2025a-CUDA-12.8.0.eb)
* JupyterHub/5.3.0-GCCcore-14.2.0 (JupyterHub-5.3.0-GCCcore-14.2.0.eb)
* ESM-2/2.0.0-foss-2025a (ESM-2-2.0.0-foss-2025a.eb)
* t-SNE-CUDA/3.0.1-foss-2025a-CUDA-12.8.0 (t-SNE-CUDA-3.0.1-foss-2025a-CUDA-12.8.0.eb)
* prompt-toolkit/3.0.52-GCCcore-14.2.0 (prompt-toolkit-3.0.52-GCCcore-14.2.0.eb)
* jupyter-bash-kernel/0.9.3-GCCcore-14.2.0 (jupyter-bash-kernel-0.9.3-GCCcore-14.2.0.eb)
* Jupyter-bundle/20250702-GCCcore-14.2.0 (Jupyter-bundle-20250702-GCCcore-14.2.0.eb)
* Qtconsole/5.7.0-GCCcore-14.2.0 (Qtconsole-5.7.0-GCCcore-14.2.0.eb)
* napari/0.6.6-foss-2025a (napari-0.6.6-foss-2025a.eb)
* ATK/2.38.0-GCCcore-14.2.0 (ATK-2.38.0-GCCcore-14.2.0.eb)
* libepoxy/1.5.10-GCCcore-14.2.0 (libepoxy-1.5.10-GCCcore-14.2.0.eb)
* GStreamer/1.26.6-GCCcore-14.2.0 (GStreamer-1.26.6-GCCcore-14.2.0.eb)
* at-spi2-core/2.56.2-GCCcore-14.2.0 (at-spi2-core-2.56.2-GCCcore-14.2.0.eb)
* at-spi2-atk/2.38.0-GCCcore-14.2.0 (at-spi2-atk-2.38.0-GCCcore-14.2.0.eb)
* GTK3/3.24.43-GCCcore-14.2.0 (GTK3-3.24.43-GCCcore-14.2.0.eb)
* Graphene/1.10.8-GCCcore-14.2.0 (Graphene-1.10.8-GCCcore-14.2.0.eb)
* shared-mime-info/2.4-GCCcore-14.2.0 (shared-mime-info-2.4-GCCcore-14.2.0.eb)
* GST-plugins-base/1.26.6-GCCcore-14.2.0 (GST-plugins-base-1.26.6-GCCcore-14.2.0.eb)
* GST-plugins-bad/1.26.6-GCCcore-14.2.0 (GST-plugins-bad-1.26.6-GCCcore-14.2.0.eb)
* PyCairo/1.28.0-GCCcore-14.2.0 (PyCairo-1.28.0-GCCcore-14.2.0.eb)
* PyGObject/3.54.2-GCCcore-14.2.0 (PyGObject-3.54.2-GCCcore-14.2.0.eb)
* libcroco/0.6.13-GCCcore-14.2.0 (libcroco-0.6.13-GCCcore-14.2.0.eb)
* cargo-c/0.10.13-GCCcore-14.2.0 (cargo-c-0.10.13-GCCcore-14.2.0.eb)
* librsvg/2.61.0-GCCcore-14.2.0 (librsvg-2.61.0-GCCcore-14.2.0.eb)
* GTK4/4.20.2-GCCcore-14.2.0 (GTK4-4.20.2-GCCcore-14.2.0.eb)
* libnotify/0.8.7-GCCcore-14.2.0 (libnotify-0.8.7-GCCcore-14.2.0.eb)
* wxWidgets/3.2.9-GCCcore-14.2.0 (wxWidgets-3.2.9-GCCcore-14.2.0.eb)
* ctffind/4.1.14-foss-2025a (ctffind-4.1.14-foss-2025a.eb)
* RELION/5.1.0-foss-2025a-CUDA-12.8.0 (RELION-5.1.0-foss-2025a-CUDA-12.8.0.eb)

[jpecar]

Yes, I chewed through most of them already and left some notes in the original relion PR. PyTorch is the obvious blocker, others are mostly manageable.

How should these cpu-only deps be processed? As a separate PR?

[Neves-P]

Perfect, thank you very much! Indeed, ideally in a separate PR or multiple PRs, depending on the size and difficulty of the installation. For example, for things like PyTorch we usually install the deps first since they usually are not as difficult, and then open a PR just for PyTorch itself.

The reason for the CPU-only separation is to make sure that those installations land in the "regular" prefix /cvmfs/software.eessi.io/versions/2023.06/software/linux/<arch>/<vendor>/<model>/software/ and don't get inadvertently placed in the dedicated accel subdir: /cvmfs/software.eessi.io/versions/2023.06/software/linux////accel/`.

To that effect, could you please also move the RELION easystack entry to a new file in the repo at easystacks/software.eessi.io/2025.06/accel/nvidia/eessi-2025.06-eb-5.4.0-2025a.yml? That way the installation will go the the accel/ subdirectory 👍

[jpecar]

CPU deps are in #1635. Lets sort those out first.

[Neves-P]

Great, thanks!