checkpoints/weights/step060000.pt: 60,000 optimizer steps in this v6 memory-modulator run. Provenance: provenance/step060000.json. - checkpoints/weights/step050000.pt: 50,000 optimizer steps in this v6 memory-modulator run. Provenance: provenance/step050000.json. - checkpoints/weights/step040000.pt: 40,000 optimizer steps in this v6 memory-modulator run. Provenance: provenance/step040000.json. - checkpoints/weights/step030000.pt: 30,000 optimizer steps in this v6 memory-modulator run. Provenance: provenance/step030000.json. checkpoints/weights/step020000.pt: 20,000 optimizer steps in v6modulator90k20261009, a planned 90k-step experiment. This is the v6 memory-modulator architecture, not v7…
Independent publisher
Junjie008
junjie008
Models in Library4
Datasets in Library0
Models on Hugging Face5
Followers—
Compare the published evidence
This profile connects the 4 model and dataset records shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Robotics.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.
Models
Intermediate inference checkpoint from the v10 two-stage run started 2026-10-15. This release contains joint step 2,000, following 1,000 steps of new-module warmup. It is not the warmup checkpoint, and no online RoboMME success rate is claimed. joint/step002000/ contains config.yaml, datasetstats.json, provenance.json, SHA256SUMS and checkpoints/weights/step002000.pt. The single.pt contains the trained MoT (including v10 modules), proprio encoder and model metadata. A separate v7/warmup adapter checkpoint is not needed. Frozen external assets such as the VAE and text encoder are not bundled; use the compatible FastWAM installation and adjust machine-specific config paths. Optimizer…