|
Download evaluation/simulation/README.md from wzlllong/value-model-benchmark-training-data: direct link, hf CLI and curl.
- Browser
- Download file 803 Bytes
-
https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/refs%2Fpr%2F5/evaluation/simulation/README.md
- Command line
-
hf download hf://datasets/wzlllong/value-model-benchmark-training-data@refs/pr/5/evaluation/simulation/README.md
-
curl -L -o README.md https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/refs%2Fpr%2F5/evaluation/simulation/README.md
803 Bytes
| # RoboValue Simulation Public Test Sets | |
| [简体中文](README_zh.md) · [Repository overview](../../README.md) | |
| Public simulation diagnostic subsets are stored here, separately from training demonstrations under `training-data/simulation/`. | |
| | Subset | Tasks | Trajectories | Time steps | Videos | Format | | |
| | --- | ---: | ---: | ---: | ---: | --- | | |
| | [Test-30](test-30/README.md) | 4 | 30 | 17,149 | 90 | Visual-input LeRobot v3.0 | | |
| [Test-30 metric coverage](test-30/README.md#metric-coverage) · [Trajectory and video index](test-30/TRAJECTORIES.md) · [Real-world Test-30](../real-world/test-30/README.md) | |
| Each released subset is an independent dataset root. Future simulation subsets can be added beneath this directory with their own data, annotations, documentation, and browsing configuration. | |