Merge pull request #322 from EPCCed/GPUnodesconfigs

Added: number of GPUs per node configurations for GPU Service upon re…

Author: agngrant

docs/services/cerebras/run.md

@@ -13,7 +13,6 @@ We also that suggest:
 - you make all paths in configuration files absolute
 - you use the `--mount_dirs` and `--python_paths` options with all `cszoo` commands, see the training example below for a full command invocation.
 
-
 For completness both the modelzoo clone and virtual environmment setup are included below:
 
 ```bash

docs/services/gpuservice/index.md

@@ -10,10 +10,10 @@ The current full specification of the EIDF GPU Service as of April 2025:
 - 59 TiB Memory
 - Ceph Persistent Volumes CephFS - up to 100TiB
 - Ceph Persistent Volumes RBD - up to 100TiB
-- 16 Nvidia H200 141GB
-- 136 Nvidia H100 80GB
-- 80 Nvidia A100 80GB
-- 56 Nvidia A100 40GB
+- 16 Nvidia H200 141GB with 8 GPUs per node
+- 152 Nvidia H100 80GB  with 8 GPUs per node
+- 80 Nvidia A100 80GB either 4 or 8 GPUs per node
+- 56 Nvidia A100 40GB with 8 GPUs per node
 - 16 Nvidia MIG A100 3G.20GB
 - 56 Nvidia MIG A100 1G.5GB