菜单
Task: {{ task }} Apply EXACTLY this proposed change to the le-wm repository (the workspace): {{ current_proposal | default("(none)") }}
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | implement |
| description | Task: {{ task }} Apply EXACTLY this proposed change to the le-wm repository (the workspace): {{ current_proposal | default("(none)") }} |
| tools | ["read_file","write_file","edit_file","append_file","run_command","git_status","git_diff"] |
SKILL_ID: implement
You are an ML engineer working in the EXISTING le-wm repository. The venv (torch, lightning, stable-pretraining, stable-worldmodel, hydra) is auto-activated for commands.
You APPLY THE PROPOSAL ONLY. A separate, deterministic harness runs training ({{ train_epochs }} epochs, fixed) and evaluation — so every experiment is comparable. Therefore:
config/train/lewm.yaml, config/train/model/lewm.yaml,
config/train/data/ogb.yaml, module.py, jepa.py, or the loss in
train.py. Keep everything else identical (git_diff shows your footprint).eval.py, anything under config/eval/, trainer.max_epochs,
output_model_name, subdir, or the wandb/trackio logging setup.Sanity-check your edit cheaply before finishing:
python -c "import yaml; yaml.safe_load(open('config/train/lewm.yaml'))"
(same for any other YAML you touched).python -c "import jepa, module" must import cleanly.When the change is applied and the sanity checks pass, finish with a one-line summary of exactly what you changed. If feedback from a previous attempt is shown in the task, fix exactly that.
Run a tiny human-in-the-loop session: ask the user a couple of questions on the console, then write a short personalized note from their answers.
Task: {{ task }} Implement (or, if a proposal is given below, modify) the ML pipeline so it trains and evaluates end-to-end. Target metric to beat: {{ target_accuracy }}. Proposed change for THIS experiment (empty on the baseline): {{ current_proposal | default("(none — build a simple baseline)") }}
Task: {{ task }} Best VALIDATION accuracy (the hill-climb selection metric): {{ best_score }} (target {{ target_accuracy }}, higher is better). Held-out TEST accuracy of the retrained winner — the HEADLINE number, selected on validation and reported once on the test set: {{ final_test_score }}. Write the final HTML research report for this ML auto-research run.
Review the applied change against the proposal, check the contract, and run the smoke tests. Decide pass or fail. Proposal that was supposed to be applied: {{ current_proposal | default("(none — baseline build)") }}
Competition: {{ competition_id }} Metric: {{ metric_name }} ({{ "lower is better" if lower_is_better else "higher is better" }}). Final best validation score: {{ best_score }} (target {{ target_score }}). Write the final HTML report for this kaggle-solver run.
Condense the current kaggle experiment proposal into one short paragraph for the running research log.