Add configurable Slurm submission workflow#1
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3869101b5a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| EXPERIMENT_CFG="${EXPERIMENTS[$TASK_ID]}" | ||
| if [[ ! -f "$EXPERIMENT_CFG" ]]; then | ||
| echo "ERROR: Experiment config not found: $EXPERIMENT_CFG" >&2 |
There was a problem hiding this comment.
The worker reads each list entry into EXPERIMENT_CFG and immediately checks -f on that raw value, but the example list uses relative paths (for example configs/experiments/...). If submission happens from a different directory (or the job starts with a different cwd), tasks fail with Experiment config not found even though the config exists under WORKSPACE_ROOT, which breaks the advertised reusable submission flow for cluster runs.
Useful? React with 👍 / 👎.
| fi | ||
|
|
||
| EXPERIMENT_LIST_FILE="$1" | ||
| WORKSPACE_ROOT="${2:-$SLURM_SUBMIT_DIR}" |
There was a problem hiding this comment.
Because set -u is enabled, WORKSPACE_ROOT="${2:-$SLURM_SUBMIT_DIR}" exits immediately when the script is invoked with only the required list argument outside a Slurm-launched environment, even though usage declares workspace_root as optional. This makes direct invocation (including local dry-run debugging) fail before any meaningful validation or execution.
Useful? React with 👍 / 👎.
1 participant
Summary
Notes