From f51d5f0296724705f2857385ae52fe2020407a5b Mon Sep 17 00:00:00 2001
From: Marcel Langer <me@sirmarcel.com>
Date: Mon, 29 Apr 2024 19:07:59 +0200
Subject: [PATCH] Add instructions to install `sphericart`

---
 gpu/sphericart/README.md | 46 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)
 create mode 100644 gpu/sphericart/README.md

diff --git a/gpu/sphericart/README.md b/gpu/sphericart/README.md
new file mode 100644
index 0000000..7380a14
--- /dev/null
+++ b/gpu/sphericart/README.md
@@ -0,0 +1,46 @@
+## Installing `sphericart` on `izar`
+
+You have to do the installation on an interactive node *with* GPU. Currently, `sphericart` compiles in parallel and uses a lot of RAM, so request a node with some amount of RAM larger than 4GB, or edit `sphericart-torch/setup.py` and remove the `--parallel` flag from the `cmake` invocation.
+
+You can request a suitable node with:
+
+```bash
+Sinteract -g gpu:1 -t 01:00:00 -m 32G
+```
+
+Once you have the node, run something along these lines:
+
+
+```bash
+
+# -- prerequisites --
+
+module purge
+module load git
+module load gcc/11.3.0
+module load python/3.10.4
+module load cuda/12.1.1
+module load cudnn/8.9.7.29-12
+
+
+# make & activate venv for this -- skip if you already have one
+cd ~
+mkdir envs
+cd envs/
+
+python -m venv torch-1
+source torch-1/bin/activate
+
+# this will install the current version of pytorch, with CUDA
+pip install torch
+
+# -- now sphericart --
+
+git clone https://github.com/lab-cosmo/sphericart
+cd sphericart
+# maybe edit setup.py if RAM is an issue
+pip install .[torch]
+
+```
+
+That's it, enjoy.