task string | task_id string | evals list | prompt string | envs list | agents list | verifiers list | num_steps int64 | steps unknown |
|---|---|---|---|---|---|---|---|---|
hello | hello/hello | [] | null | [] | [] | [
"hello"
] | 3 | [
{
"id": "box",
"type": "deploy_sandbox",
"sandbox_name": "box",
"sandbox_mode": "vm",
"sandbox_type": "local"
},
{
"id": "load",
"type": "load_artifact",
"sandbox_name": "box",
"artifact_id": "greeting",
"destination_path": "/app/greeting"
},
{
"id": "hello",
... |
agentenv-hf test dataset
The smallest bundle: load a greeting into a local sandbox and check it.
An AgentEnv bundle with its recorded runs, published with agentenv-hf.
Tables
hello_tasks: one row per task of the bundle, with its prompt, envs, verifiers and steps.hello_episodes: one row per recorded run, with its status, model, reward and scores, the prompt and response, and the agent's transcript as chatmessageswhen its trajectory format is one agentenv-hf reads.
raw/hello.jsonl holds each run's record and native trajectory, and bundles/hello/ the bundle itself.
This is a test dataset for agentenv-hf, published with
agent-env hf publish. Each name's license and sources are in its own section below.
Run it
pip install agentenv-framework huggingface_hub
hf download earakely-scale/agentenv-hf-smoke --repo-type dataset --local-dir agentenv-dataset
agent-env run agentenv-dataset/bundles/hello
hello
The AgentEnv Framework's built-in example bundle (Apache-2.0).
v3
A recorded run of the v3 marine-port weeks of agentenv-portsim, built on PortSimEnv by Adithya S Kolavi. Its tasks and transcripts are CC BY-SA 4.0 and contain data from the Port de Barcelona open data portal. The maintained PortSim dataset is earakely-scale/PortSimEnv-AgentEnv.
- Downloads last month
- 6