WorldPilot LIBERO: Precomputed Cache for AI Training Pipeline

A precomputed Cosmos cache used by the public WorldPilot LIBERO training pipeline. The cache was released by author Chedan86 and hosted on Hugging Face, with a last update timestamp of 2026-06-08 15:03:50. It serves as a foundational resource for the WorldPilot project's training infrastructure.

Use Cases

• Reproduce training runs for the WorldPilot LIBERO model based on the provided precomputed cache.
• Benchmark AI training pipeline performance using the released cache data.
• Analyze the structure and content of a Cosmos cache used in a major AI training project.

Strengths

• Directly supports a public, documented training pipeline (WorldPilot LIBERO).
• Associated with a specific model checkpoint release and project codebase.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Chedan86 on Hugging Face, part of the WorldPilot project.

Collection Method

Precomputed for the LIBERO training pipeline.

Freshness

Last updated 2026-06-08 15:03:50; freshness should be verified.