Automation: OpenEnv Examples Automation ID: openenv-examples Automation memory: $CODEX_HOME/automations/openenv-examples/memory.md
Run and expand on the examples to test and improve them.
Take the most recent three examples from the examples directory that have not been run based on memory and run them on a real-world workload.
TASK:
Automation: Post-training agent experiments Automation ID: post-training-agent-experiments Automation memory: $CODEX_HOME/automations/post-training-agent-experiments/memory.md Last run: 2026-06-23T05:01:48.919Z (1782190908919)
Run the Training Agents Terminal-Bench loop.
GOAL Train an approximately 2B parameter open model to exceed 40 on Terminal-Bench. If full Terminal-Bench scoring is blocked, use the strongest available proxy gate, such as TB Lite, Harbor, Inspect, or an OpenEnv tbench2-style held-out subset, but label it as a proxy and do not claim a full benchmark win.
| #!/usr/bin/env python3 | |
| # /// script | |
| # requires-python = ">=3.10,<3.13" | |
| # dependencies = [ | |
| # "openenv @ git+https://github.com/burtenshaw/OpenEnv.git@ben/hf-sandbox-provider", | |
| # "openenv-browsergym-env @ git+https://huggingface.co/spaces/openenv/browsergym_env", | |
| # "huggingface_hub @ git+https://github.com/huggingface/huggingface_hub.git@ben/sandbox-pool-serve", | |
| # "torch", | |
| # "transformers>=5.0.0", | |
| # "peft", |
| # /// script | |
| # dependencies = [ | |
| # "datasets>=3.0.0", | |
| # "peft>=0.13.0", | |
| # "torch", | |
| # "transformers", | |
| # "trl>=1.5.0", | |
| # ] | |
| # /// |
| import json | |
| import os | |
| import re | |
| import shlex | |
| import subprocess | |
| import sys | |
| import time | |
| import traceback | |
| import urllib.error | |
| import urllib.request |
| #!/usr/bin/env python3 | |
| """ | |
| Test script for benchmark environment concurrency. | |
| Run the server first: | |
| cd benchmark && uvicorn server.app:app --reload --port 8000 | |
| Then run this script: | |
| python test_concurrency.py --requests 10 --wait 1.0 | |
| """ |
| # /// script | |
| # requires-python = ">=3.13" | |
| # dependencies = [ | |
| # "huggingface-hub>=1.1.4", | |
| # "python-dotenv>=1.2.1", | |
| # "pyyaml>=6.0.3", | |
| # "requests>=2.32.5", | |
| # ] | |
| # /// |
| #!/usr/bin/env python3 | |
| """ | |
| GRPO training for BrowserGym tasks using TRL's `GRPOTrainer` and the OpenEnv HTTP | |
| environment interface. | |
| """ | |
| from __future__ import annotations | |
| from datetime import datetime | |
| import sys |
| #!/usr/bin/env python3 | |
| # | |
| # /// script | |
| # requires-python = ">=3.12" | |
| # dependencies = [ | |
| # "git+https://github.com/huggingface/trl.git#egg=trl[vllm]", | |
| # "trackio==0.8.1", | |
| # /// | |
| """ |
| #!/usr/bin/env python3 | |
| # | |
| # /// script | |
| # requires-python = ">=3.12" | |
| # dependencies = [ | |
| # "git+https://github.com/huggingface/trl.git#egg=trl[vllm]", | |
| # "trackio==0.8.1", | |
| # /// | |
| """ |