songminjae/vjepa-dpxpsi-b1k-t15

robotics model by songminjae

Older snapshot: check the original source for current information.

From the publisher

Model card & documentation

Source preview

Read the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.

Action policy checkpoints for the BEHAVIOR-1K R1Pro robot. A frozen V-JEPA 2.1 ViT-L/16 (384 px) vision encoder and a frozen Qwen3-0.6B text encoder feed a Psi0-style MM-DiT action expert (vjepapolicypsi, formerly dpxpsi) that generates 24-step action chunks by flow matching. All three checkpoints come from one training run on 15 BEHAVIOR-1K tasks (3 epochs = 155,211 steps planned; stopped at step 116,400). Code: https://github.com/songminjae/vjepa-based-policy, branch t15-wrist-psi-renewal (commit 143cc3c). Branch t15-wrist-psi (commit 4e2d676) holds the same model code under the old name dpxpsi. Each file is a torch.save dict with model (raw weights), ema (EMA weights, used for inference), cfg (model config), args (training arguments), arch, step and normstatssha. None of them holds optimizer state. ckpt052500.pt stores arch dpxpsi, the other two vjepapolicypsi; the renewal branch reads both names. normstatstrain.json holds the action / state normalisation statistics (absolute action mode); its sha (e32883d4e0042aaf) must match the checkpoint's normstatssha. args.json holds the…

songminjae/vjepa-dpxpsi-b1k-t15 on Hugging Face A short preview, not the full document.

Read the full model card ↗ · Preview checked 2026-10-31T03:04:07.947Z

Inside the original model card — Document outline

Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.

Documentation belongs to its respective authors. Reported project/model license: other. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.

Model overview

songminjae/vjepa-dpxpsi-b1k-t15 is a robotics model repository published by songminjae on Hugging Face. The source reports the pytorch library.

This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.

Model facts

Task: robotics
Library: pytorch
Recent downloads (30 days): 0
Cumulative likes: 0
Hugging Face trending score: 0
Reported safetensors parameters: Not reported
Architecture: Not reported
License: other
Access: Not gated by Hugging Face
Created: 2026-10-26T06:26:06.000Z
Last modified: 2026-10-26T06:27:37.000Z

Compatibility and lineage

Language coverage was not reported.

Base-model lineage was not included.

Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.

Use and evaluate

Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.

No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.

Model card and usage instructions

Model files and configuration

Community discussion