paper2code

分类: 开发与编程 | 上传者: PrathamLearnsToCode | 下载: 0 | 版本: v1.0（最新）

将 arxiv 论文转换为最小的、基于引文的 Python 实现。当用户使用 arxiv URL 或论文 ID 运行 /paper2code、说“实现本文”或粘贴请求实现的 arxiv 链接时触发。诚实地标记所有含糊之处。绝不发明论文中未提及的实现细节。

更新日志: Source: GitHub https://github.com/PrathamLearnsToCode/paper2code

目录结构

当前层级: tree/main/skills/paper2code/

📁 guardrails/
- 📄 badly_written_papers.md 8.8 KB
- 📄 hallucination_prevention.md 8.1 KB
- 📄 scope_enforcement.md 7.9 KB
📁 knowledge/
- 📄 loss_functions.md 9.8 KB
- 📄 paper_to_code_mistakes.md 10.3 KB
- 📄 training_recipes.md 9.1 KB
- 📄 transformer_components.md 14.0 KB
📁 pipeline/
- 📄 01_paper_acquisition.md 7.0 KB
- 📄 02_contribution_identification.md 8.4 KB
- 📄 03_ambiguity_audit.md 10.5 KB
- 📄 04_code_generation.md 11.1 KB
- 📄 05_walkthrough_notebook.md 6.8 KB
📁 scaffolds/
- 📄 config_template.yaml 2.6 KB
- 📄 data_template.py 3.2 KB
- 📄 evaluate_template.py 1.7 KB
- 📄 loss_template.py 1.4 KB
- 📄 model_template.py 4.1 KB
- 📄 readme_template.md 1.8 KB
- 📄 reproduction_notes_template.md 3.5 KB
- 📄 train_template.py 5.7 KB
📁 scripts/
- 📄 extract_structure.py 11.0 KB
- 📄 fetch_paper.py 15.4 KB
📁 worked/
- 📁 attention_is_all_you_need/
  - 📁 configs/
    
    📄 base.yaml 3.4 KB
  - 📁 notebooks/
    
    📄 walkthrough.ipynb 26.9 KB
  - 📁 src/
    
    📄 data.py 5.9 KB
    
    📄 evaluate.py 2.2 KB
    
    📄 loss.py 2.9 KB
    
    📄 model.py 22.6 KB
    
    📄 train.py 3.4 KB
    
    📄 utils.py 3.1 KB
  - 📄 README.md 2.6 KB
  - 📄 REPRODUCTION_NOTES.md 5.9 KB
  - 📄 requirements.txt 39 B
  - 📄 review.md 5.1 KB
- 📁 ddpm/
  - 📁 configs/
    
    📄 base.yaml 3.8 KB
  - 📁 notebooks/
    
    📄 walkthrough.ipynb 13.7 KB
  - 📁 src/
    
    📄 data.py 3.2 KB
    
    📄 evaluate.py 8.0 KB
    
    📄 loss.py 1.9 KB
    
    📄 model.py 14.5 KB
    
    📄 train.py 6.5 KB
    
    📄 utils.py 8.0 KB
  - 📄 README.md 2.8 KB
  - 📄 REPRODUCTION_NOTES.md 6.1 KB
  - 📄 requirements.txt 39 B
  - 📄 review.md 4.4 KB
📄 SKILL.md 5.2 KB

SKILL.md

---
name: paper2code
description: Converts an arxiv paper into a minimal, citation-anchored Python implementation. Trigger when user runs /paper2code with an arxiv URL or paper ID, says "implement this paper", or pastes an arxiv link asking for implementation. Flags all ambiguities honestly. Never invents implementation details not stated in the paper.
---

# paper2code — Orchestration

You are executing the paper2code skill. This file governs the high-level flow. Each stage dispatches to a detailed reasoning protocol in `pipeline/`. Do NOT skip stages. Do NOT combine stages. Execute them in order.

## Parse arguments

Extract from the user's input:
- `ARXIV_ID`: the arxiv paper ID (e.g., `2106.09685`). Strip any URL prefix.
- `MODE`: one of `minimal` (default), `full`, `educational`.
- `FRAMEWORK`: one of `pytorch` (default), `jax`, `numpy`.

If the user provided a full URL like `https://arxiv.org/abs/2106.09685`, extract the ID `2106.09685`.
If the user provided a versioned ID like `2106.09685v2`, keep the version.

## Set up working directory

Create a temporary working directory: `.paper2code_work/{ARXIV_ID}/`
This is where intermediate artifacts go. The final output goes in the current directory under `{paper_slug}/`.

## Install dependencies

Run via Bash:
```bash
pip install pymupdf4llm pdfplumber requests pyyaml
```

## Execute pipeline

### Stage 1 — Paper Acquisition and Parsing
Read and follow: `pipeline/01_paper_acquisition.md`

Run the helper script to fetch and parse the paper:
```bash
python skills/paper2code/scripts/fetch_paper.py {ARXIV_ID} .paper2code_work/{ARXIV_ID}/
```
Then run structure extraction:
```bash
python skills/paper2code/scripts/extract_structure.py .paper2code_work/{ARXIV_ID}/paper_text.md .paper2code_work/{ARXIV_ID}/
```
Verify the outputs exist before proceeding. If extraction failed, follow the fallback protocol in `pipeline/01_paper_acquisition.md`.

The script also searches for official code repositories (in the paper text and on the arxiv page) and saves any found links to `paper_metadata.json` under the `official_code` key. Verify these links before relying on them — see Step 8 in `pipeline/01_paper_acquisition.md`.

### Stage 2 — Contribution Identification
Read and follow: `pipeline/02_contribution_identification.md`

Read the parsed paper sections. Identify the single core contribution. Classify the paper type. Write the contribution statement. Save it to `.paper2code_work/{ARXIV_ID}/contribution.md`.

### Stage 3 — Ambiguity Audit
Read and follow: `pipeline/03_ambiguity_audit.md`

Before reading this stage, also read: `guardrails/hallucination_prevention.md`

Go through every implementation-relevant detail. Classify each as SPECIFIED, PARTIALLY_SPECIFIED, or UNSPECIFIED. Save the audit to `.paper2code_work/{ARXIV_ID}/ambiguity_audit.md`.

### Stage 4 — Code Generation
Read and follow: `pipeline/04_code_generation.md`

Before writing code, read:
- `guardrails/scope_enforcement.md` — to determine what's in and out of scope
- `guardrails/badly_written_papers.md` — if the paper is vague or inconsistent
- The relevant knowledge files in `knowledge/` for the paper's domain
- The scaffold templates in `scaffolds/` for the expected file structure

Determine the `paper_slug` from the paper title (lowercase, underscores, no special chars).
Generate all files under `{paper_slug}/` in the current working directory.

### Stage 5 — Walkthrough Notebook
Read and follow: `pipeline/05_walkthrough_notebook.md`

Generate the walkthrough notebook that connects paper sections to code with runnable sanity checks. Save to `{paper_slug}/notebooks/walkthrough.ipynb`.

## Cleanup

Remove the `.paper2code_work/` directory after successful completion.

## Final output

Print a summary:
```
✓ paper2code complete for: {paper_title}
  Output directory: {paper_slug}/
  Files generated: {list of files}
  Unspecified choices: {count} (see REPRODUCTION_NOTES.md)
  Mode: {MODE} | Framework: {FRAMEWORK}
```

## Mode-specific behavior

- **minimal** (default): Core contribution only. Training loop only if contribution involves training. No data pipeline beyond Dataset skeleton.
- **full**: Core contribution + full training loop + data pipeline + evaluation pipeline. More code, same citation rigor.
- **educational**: Same as minimal but with extra inline comments explaining ML concepts, expanded walkthrough notebook with theory sections, and a `PAPER_GUIDE.md` that walks through the paper section by section.

## Guardrails — always active

These apply at ALL stages. Read them if you haven't already:
- `guardrails/hallucination_prevention.md` — the most important file in this skill
- `guardrails/scope_enforcement.md` — what to implement and what to skip
- `guardrails/badly_written_papers.md` — what to do when the paper is unclear

## Knowledge base — consult as needed

Before implementing any of these components, read the corresponding knowledge file:
- Transformer layers, attention, positional encoding → `knowledge/transformer_components.md`
- Optimizers, LR schedules, batch size semantics → `knowledge/training_recipes.md`
- Cross-entropy, contrastive loss, diffusion loss, ELBO → `knowledge/loss_functions.md`
- Framework-specific pitfalls, notation mismatches → `knowledge/paper_to_code_mistakes.md`

登录后下载/点赞/收藏 ❤ 39 | ★ 0

paper2code

目录结构

SKILL.md

举报内容

提示