hamishivi/agent-task-terminal-lego-15k

Terminal-Lego-15k for tmax

Images require building: the dataset conversion and complete task archive are published. 47 task image(s) were published during preparation; the remaining images are intentionally left for the included resumable build script. No build job is running for this release. Only tasks with available images can run immediately. See image-build-status.json for the publication-time snapshot.

All 15,048 task directories from Lego-X/Terminal-Lego-15k, pinned to 9c197f1c2e87b64cc316b1a5bfcef57b584929f0, converted to tmax's swerl_vanillux_sandbox format.

The train split uses the same messages, ground_truth, dataset, env_config, and source schema as the other hamishivi/agent-task-* datasets. Messages use the tmax Vanillux templates; dataset is passthrough. Task IDs are prefixed with terminal_lego__ to avoid collisions when combining task archives. task-manifest.json maps them back to original paths.

task-data.tar.gz contains a directory per task with all original task files preserved byte-for-byte, plus image.txt. Instructions, verifiers, build contexts, metadata, and reference solutions are retained. The tmax loader exposes only environment seed files and defers tests until submission; reference solutions are not mounted into the agent sandbox.

Usage

from datasets import load_dataset
from huggingface_hub import hf_hub_download

train = load_dataset("hamishivi/agent-task-terminal-lego-15k", split="train")
archive = hf_hub_download("hamishivi/agent-task-terminal-lego-15k", "task-data.tar.gz", repo_type="dataset")

Extract the archive and set the sandbox task_data_dir to the directory containing the task directories. Every row specifies its image explicitly. Task resource requirements and verifier timeouts remain in the original task.toml; configure the training harness accordingly.

Images

Images use hamishi740/agent-task-terminal-lego-15k:<source-environment-hash> and target Linux AMD64. See task-manifest.json for build status.

Validation and attribution

All rows passed schema and Parquet round-trip checks, and every archived task file was checked against the downloaded source bytes. See validation.json. This does not constitute an evaluation of every task or a full training run.

The upstream license is apache-2.0. Consult the upstream dataset card for citation and provenance. tmax Vanillux prompts are adapted from mini-swe-agent (MIT).

Resume image builds

Requires Docker Buildx and a Docker Hub login with write access to the hamishi740 target repositories. The script skips tags already published, retains per-image logs, and can be rerun after failures.

mkdir -p tasks
tar -xzf task-data.tar.gz -C tasks
python build_dataset_images.py --task-data-dir tasks --manifest task-manifest.json --workers 8 --timeout 1800

To use another Docker Hub namespace, update the image references consistently in the manifest, training records, and task image.txt files before building.

Packaging repair

7,518 upstream Dockerfiles copy environment/task_file even though the directory is absent from Git. Empty directories were restored for those tasks and included in the archive; no input files were invented or source file contents changed. These repairs are listed in task-manifest.json.

Sample runtime checks

Sample task_00000 scored 0 with no solution. Its upstream oracle initially failed because /app/task_file/output was absent; creating that directory before running the unchanged oracle produced reward 1. The conversion retains the upstream oracle unchanged. See smoke-results.json.

Apptainer images

The current Apptainer pool and unified download manifest are maintained in TMaxxx/agent-task-terminal-lego-15k. New SIF uploads go to TMaxxx; earlier images remain available here. Use the downloader and manifest in the linked repository to retrieve all available images with tmax-compatible filenames.