|
Download README.md from ArchCoder/Openenv: direct link, hf CLI and curl.
- Browser
- Download file 2.37 kB
-
https://huggingface.co/spaces/ArchCoder/Openenv/resolve/fb9822bd288e075643b09163ccbced7b0affe4fa/README.md
- Command line
-
hf download hf://spaces/ArchCoder/Openenv@fb9822bd288e075643b09163ccbced7b0affe4fa/README.md
-
curl -L -o README.md https://huggingface.co/spaces/ArchCoder/Openenv/resolve/fb9822bd288e075643b09163ccbced7b0affe4fa/README.md
2.37 kB
| title: PyTorch Debug Env | |
| emoji: π₯ | |
| colorFrom: yellow | |
| colorTo: red | |
| sdk: docker | |
| app_port: 7860 | |
| short_description: RL environment for diagnosing broken PyTorch training jobs | |
| tags: | |
| - openenv | |
| - pytorch | |
| - reinforcement-learning | |
| - debugging | |
| - ml-training | |
| - agent | |
| pinned: true | |
| # PyTorch Debug Env π₯ | |
| A complete [OpenEnv](https://meta-pytorch.org/OpenEnv/) environment for the **Meta PyTorch Hackathon** where an AI agent investigates and diagnoses broken PyTorch training jobs. | |
| ## Quick Start | |
| ```python | |
| from openenv import AutoEnv, AutoAction | |
| env = AutoEnv.from_env("ArchCoder/pytorch-debug-env") | |
| Action = AutoAction.from_env("ArchCoder/pytorch-debug-env") | |
| with env.sync() as client: | |
| result = client.reset(task_id="easy") | |
| action = Action( | |
| current_hypothesis={ | |
| "bug_type": "missing_zero_grad", | |
| "affected_file": "train.py", | |
| "confidence": 0.7 | |
| }, | |
| commit_diagnosis=False | |
| ) | |
| step_result = client.step(action) | |
| ``` | |
| ## API Endpoints | |
| | Endpoint | Method | Description | | |
| |----------|--------|-------------| | |
| | `/` | GET | Environment info | | |
| | `/health` | GET | Health check | | |
| | `/reset?task_id=easy` | POST | Start new episode | | |
| | `/step` | POST | Submit hypothesis + action | | |
| | `/state` | GET | Current episode state | | |
| ## Tasks | |
| | Task | Difficulty | Description | | |
| |------|-----------|-------------| | |
| | `easy` | β | Single-file bug β missing `zero_grad`, wrong loss | | |
| | `medium` | ββ | Multi-file root cause β data leakage, scheduler mismatch | | |
| | `hard` | βββ | Silent failure β memory leak, AMP overflow, red herrings | | |
| ## Reward Structure | |
| - **Hypothesis delta** (60%) β reward for improving your bug hypothesis each step | |
| - **Investigation** (20%) β reward for inspecting the right files | |
| - **Final diagnosis** (20%) β accuracy of committed diagnosis vs ground truth | |
| Scores range from `0.0` to `1.0`. Partial credit for correct bug category on hard tasks. | |
| ## Environment State | |
| Each episode provides a synthetic PyTorch repo with: | |
| - Source files (`train.py`, `model/`, `data/`, `config/`) | |
| - Loss curves and GPU memory profiles | |
| - Training logs with realistic noise and red herrings | |
| The agent reveals files progressively across up to 5β6 steps, refining its hypothesis before committing a final diagnosis. | |
| ## Author | |
| **Priyansh Saxena** β IIIT Gwalior |