# Log index — Hugging Face, Zenodo, MD production, automation

All paths below are relative to the **tubulin-cppf-md** repository root unless noted. Logs are **append-only operational records** (upload progress, `gmx`/`mdrun` launches, polling scripts). They may contain **local paths** and **timestamps** from the HPC; they should **not** contain secrets if you followed `docs/HUGGINGFACE_DATASET.md` (never paste tokens into logs or chat).

## Mirror upload (`revision_exec` tree, minus prod trajectories)

Script: `revision_exec/scripts/huggingface_upload_revision_exec_mirror.sh` — uploads `revision_exec/**` to the dataset with preserved structure, but **excludes `**/prod/**/*.xtc`** so files already published via `huggingface_upload_trajectories.sh` are not duplicated. Log your run to a new `revision_exec/logs/hf_mirror_revision_exec_YYYYMMDD.nohup.log` and add a one-line note below.

## Revision export (`run_export_all.sh`, Step 2 xvg)

Live logs live under **`revision_exec/analysis_revision/work/`**, which is **gitignored** (large/clean tree). To **preserve progress in git**, periodically copy the main log into `revision_exec/logs/`:

```bash
TS=$(date +%Y%m%d_%H%M%S)
cp -a revision_exec/analysis_revision/work/export_all_9sys.log \
  "revision_exec/logs/analysis_revision_export_all_9sys_SNAPSHOT_${TS}.log"
```

| File | Notes |
|------|--------|
| `revision_exec/logs/analysis_revision_export_all_9sys_SNAPSHOT_20260504_215951.log` | In-progress nine-system export (includes `dimer_rep1` complete, `dimer_rep2` hbond running as of snapshot time). |
| `revision_exec/logs/analysis_revision_dimer_rep1_export_SNAPSHOT_20260504_215951.log` | Earlier rep1-only export attempt log. |
| `revision_exec/logs/analysis_revision_dimer_rep1_rerun_SNAPSHOT_20260504_215951.log` | Rep1 rerun log. |

After the job finishes, take **one final snapshot** with the same `cp -a` command and commit. For Hub archival of the same files without committing huge history, use `huggingface_upload_revision_exec_mirror.sh` (the `work/` tree uploads under `revision_exec/analysis_revision/work/` on the dataset).

## Hugging Face Hub (`hf_*`)

| File | Approx. size (order of magnitude) | Contents |
|------|-------------------------------------|----------|
| `revision_exec/logs/hf_upload_nohup.log` | Large (~10 MB+) | Main batch upload under `nohup`; commits per file, retries, hub API chatter. |
| `revision_exec/logs/hf_upload_dimer_extensions_20260504.nohup.log` | ~1–2 MB | Dimer extension segments upload session. |
| `revision_exec/logs/hf_upload_dimer_rep2_part0003_20260502.nohup.log` | sub‑MB | Rep2 part file upload slice. |
| `revision_exec/logs/hf_upload_monomer_beta_rep2_20260501.nohup.log` | sub‑MB | Monomer beta rep2 upload. |
| `revision_exec/logs/hf_poll_upload_incremental_monomers.nohup.log` | ~1–2 MB | Incremental monomer polling / upload helper (`hf_poll_upload_incremental_monomers` workflow). |

**Related repo artefacts (not in this folder)**

- `revision_exec/scripts/huggingface_upload_trajectories.sh` — canonical upload driver (`--dry-run`, `--bundle`, `--create-repo`).
- `revision_exec/scripts/HF_DATASET_REPO.env.example` — template for `HF_DATASET_REPO`.
- `revision_exec/HF_UPLOAD_SHA256SUMS_*.txt` — checksum manifests to ship alongside the Hub dataset and for Zenodo cross-reference.

## Zenodo

| File | Contents |
|------|----------|
| `revision_exec/logs/zenodo_upload_nohup.log` | Zenodo-related upload or staging log (smaller subset / metadata strategy per project policy). |

See `docs/ZENODO_UPLOAD_SERVER.md` when using Zenodo in parallel with Hugging Face.

## MD production and GPU queue (`rep*`, `monomer_*`, `auto_*`, `dimer_queue_*`)

These logs capture **mdrun** / extension jobs and small **shell wrappers** that chain jobs after checkpoints:

- `rep{1,2,3}_gpu*_extend*.nohup.log` — extension runs per replicate and GPU slot.
- `monomer_*_gpu*_200ns.nohup.log` — monomer 200 ns production windows.
- `auto_start_*.log` / `auto_start_*.sh` — automated starters after a stage completes.
- `auto_after_*.log` / `auto_after_*.sh` — chained “next segment” automation.
- `dimer_queue_*.log` — dimer job queue coordinator.
- `*.pid` — process id sidecars for the corresponding nohup jobs.

Use them to answer “**which GPU / which day / which extension part** produced this `xtc`?” without inferring from file dates alone.

## Other

| File | Contents |
|------|----------|
| `revision_exec/logs/complex_validation_report.txt` | Validation report (full). |
| `revision_exec/logs/complex_validation_report_clean.txt` | Cleaned / filtered variant for sharing. |
| `revision_exec/logs/gpuooo_6000q3.nohup.log` | Node- or queue-specific session. |

## Maintenance

- When starting a new long upload: `nohup bash revision_exec/scripts/huggingface_upload_trajectories.sh ... > revision_exec/logs/hf_upload_YYYYMMDD_description.nohup.log 2>&1 &`
- Add a one-line row to the **Hugging Face** table above so future readers find the new file quickly.