| ## GitHub Project board (ready to use) | |
| ### Columns | |
| 1) Backlog | |
| 2) Ready | |
| 3) In Progress | |
| 4) Review | |
| 5) Done | |
| ### Labels to create | |
| `data` `modeling` `evaluation` `demo` `docs` `infra` | |
| ### Issues (create and assign) | |
| W1 | |
| - `DATA-1` Collect Reddit Jokes CSV and document source (label: data) | |
| - `DATA-2` Implement cleaning + filters in `download_data.py` (label: data) | |
| - `DATA-3` Train/val/test split + JSONL export (label: data) | |
| - `DATA-4` Dataset stats + summary (label: data, docs) | |
| - `DOC-1` Update README with pipeline usage (label: docs) | |
| W2 | |
| - `EVAL-1` Define fixed test prompts (50+) (label: evaluation) | |
| - `EVAL-2` Draft human rubric (coherence, creativity, faithfulness) (label: evaluation) | |
| - `BASE-1` Baseline generation script (distilgpt2) (label: modeling) | |
| - `BASE-2` Baseline failure analysis examples (label: evaluation) | |
| W3 | |
| - `TRAIN-1` Fine-tuning script (HF Trainer) (label: modeling) | |
| - `TRAIN-2` Track training curves + save checkpoint (label: modeling) | |
| W4 | |
| - `CTRL-1` Add control tokens or fields (level/setting/tone) (label: modeling) | |
| - `EVAL-3` Baseline vs fine-tuned evaluation report (label: evaluation) | |
| - `SAFE-1` Add basic safety/length filters (label: evaluation) | |
| W5 | |
| - `DEMO-1` Build Streamlit/Gradio demo (label: demo) | |
| - `DOC-2` Final report (4-8 pages) (label: docs) | |
| - `DOC-3` Presentation slides (label: docs) | |
| ### Suggested board setup | |
| - Put all W1 items into Backlog | |
| - Move current week items to Ready | |
| - Only move to Review after a PR is open | |