Skip to content

Latest commit

 

History

History
165 lines (139 loc) · 8.02 KB

MODEL_CARD.md

File metadata and controls

165 lines (139 loc) · 8.02 KB

Model Card — TRON2 RL deployment policies

Status: every entry below is a template with ⚠ TO CONFIRM placeholders. The model owner and legal must complete each field before the first public release. Do not cut a public tag while any ⚠ TO CONFIRM remains.

This document covers the six ONNX weight files checked into controllers/model/ in this repository. See THIRD_PARTY_NOTICES.md §2 for the license-status summary; this file is the operational / behavioral model card.

Model index

ID Path Size (bytes) SHA-256 Role
SF_TRON2A/policy.onnx controllers/model/SF_TRON2A/policy.onnx 791 050 0b353a087c912c33b9ba690560f3501cf7bf2bf25fde91c07ee2bdfb36502d3a Policy network for SF_TRON2A (sole-ankle biped)
SF_TRON2A/encoder.onnx controllers/model/SF_TRON2A/encoder.onnx 588 998 5f7e2b8865fda7c284f0dd98b79f5c1d78935c83cbf43bf311d768154937111e Observation encoder for SF_TRON2A
WF_TRON2A/policy.onnx controllers/model/WF_TRON2A/policy.onnx 770 148 3000df452681056738a15b46fa67f4f8436b34bbb6dcc6b22fa08b1b1f8dd071 Policy network for WF_TRON2A (wheeled-foot biped)
WF_TRON2A/encoder.onnx controllers/model/WF_TRON2A/encoder.onnx 503 276 507d0630d78873f7aabfeab4eae9d7669610d709fcc903c4296d1908da54b3e7 Observation encoder for WF_TRON2A
DASF_TRON2A/policy.onnx controllers/model/DASF_TRON2A/policy.onnx 863 234 473bae82c4b09420f1013c37b234ab7475c0183b0e7f79df50c72f744b5fa0ca Policy network for DASF_TRON2A (dual-arm sole-foot humanoid)
DASF_TRON2A/encoder.onnx controllers/model/DASF_TRON2A/encoder.onnx 852 848 31ff1f3f756298421c3d88e075103e5e9ada560573200f1ab42644ca7d14be51 Observation encoder for DASF_TRON2A

SF/WF SHA-256 values were recorded 2026-07-16. Those four blobs are byte-identical to tron2-rl-deploy-ros/tron2_controllers/config/{SF,WF}_TRON2A/policy/{policy,encoder}.onnx in the sibling repository — this file and that repo's THIRD_PARTY_NOTICES.md §3 must be updated together.

DASF size and SHA-256 values were recorded from the working tree on 2026-08-20. No sibling-repository equivalence or training provenance is asserted for those two files.

Consumed by:

  • SF_TRON2A/* → controllers/SolefootController.py
  • WF_TRON2A/* → controllers/WheelfootController.py
  • DASF_TRON2A/* → controllers/DASFController.py

SF_TRON2A/policy.onnx

  • Path: controllers/model/SF_TRON2A/policy.onnx
  • Size / SHA-256: 791 050 B — 0b353a087c912c33b9ba690560f3501cf7bf2bf25fde91c07ee2bdfb36502d3a
  • Checkpoint id / hash: ⚠ TO CONFIRM (map SHA-256 above to the internal training-run checkpoint id)
  • Training run (framework, commit, date): ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM (simulator + domain randomization ranges, or real-world logs, or both — with data license status)
  • Evaluation: ⚠ TO CONFIRM (sim benchmark; real-hardware velocity tracking, joint-limit safety, thermal envelope)
  • Intended use: locomotion control for the SF_TRON2A (sole-ankle) variant, driven at the tick rate configured in controllers/model/SF_TRON2A/params.yaml, on a robot suspended / mounted for initial bring-up.
  • Out-of-scope use: any hardware variant other than SF_TRON2A; any operating envelope outside the training / evaluation distribution; unsuspended operation before the operator has confirmed the target behavior.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM (Apache-2.0 with the code / separate license / controlled external download only)

SF_TRON2A/encoder.onnx

  • Path: controllers/model/SF_TRON2A/encoder.onnx
  • Size / SHA-256: 588 998 B — 5f7e2b8865fda7c284f0dd98b79f5c1d78935c83cbf43bf311d768154937111e
  • Checkpoint id / hash: ⚠ TO CONFIRM
  • Training run: ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM
  • Evaluation: ⚠ TO CONFIRM
  • Intended use: observation encoding paired with the SF policy above; not intended to be used with any other policy.
  • Out-of-scope use: as above.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM

WF_TRON2A/policy.onnx

  • Path: controllers/model/WF_TRON2A/policy.onnx
  • Size / SHA-256: 770 148 B — 3000df452681056738a15b46fa67f4f8436b34bbb6dcc6b22fa08b1b1f8dd071
  • Checkpoint id / hash: ⚠ TO CONFIRM
  • Training run: ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM
  • Evaluation: ⚠ TO CONFIRM
  • Intended use: locomotion control for the WF_TRON2A (wheeled-foot) variant, driven at the tick rate configured in controllers/model/WF_TRON2A/params.yaml, on a robot suspended / mounted for initial bring-up.
  • Out-of-scope use: any hardware variant other than WF_TRON2A; operating envelope outside the training / evaluation distribution; unsuspended operation before the operator has confirmed behavior.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM

WF_TRON2A/encoder.onnx

  • Path: controllers/model/WF_TRON2A/encoder.onnx
  • Size / SHA-256: 503 276 B — 507d0630d78873f7aabfeab4eae9d7669610d709fcc903c4296d1908da54b3e7
  • Checkpoint id / hash: ⚠ TO CONFIRM
  • Training run: ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM
  • Evaluation: ⚠ TO CONFIRM
  • Intended use: observation encoding paired with the WF policy above; not intended to be used with any other policy.
  • Out-of-scope use: as above.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM

DASF_TRON2A/policy.onnx

  • Path: controllers/model/DASF_TRON2A/policy.onnx
  • Size / SHA-256: 863 234 B — 473bae82c4b09420f1013c37b234ab7475c0183b0e7f79df50c72f744b5fa0ca
  • Checkpoint id / hash: ⚠ TO CONFIRM (map SHA-256 above to the internal training-run checkpoint id)
  • Training run (framework, commit, date): ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM (simulator + domain randomization ranges, or real-world logs, or both — with data license status)
  • Evaluation: ⚠ TO CONFIRM (sim benchmark; real-hardware velocity tracking, joint-limit safety, thermal envelope, and upper/lower-body coordination)
  • Intended use: locomotion control for the DASF_TRON2A dual-arm sole-foot variant, driven at the tick rate configured in controllers/model/DASF_TRON2A/params.yaml, on a robot suspended / mounted for initial bring-up.
  • Out-of-scope use: any hardware variant other than DASF_TRON2A; any operating envelope outside the training / evaluation distribution; unsuspended operation before the operator has confirmed the target behavior.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM (Apache-2.0 with the code / separate license / controlled external download only)

DASF_TRON2A/encoder.onnx

  • Path: controllers/model/DASF_TRON2A/encoder.onnx
  • Size / SHA-256: 852 848 B — 31ff1f3f756298421c3d88e075103e5e9ada560573200f1ab42644ca7d14be51
  • Checkpoint id / hash: ⚠ TO CONFIRM
  • Training run: ⚠ TO CONFIRM
  • Training data description: ⚠ TO CONFIRM
  • Evaluation: ⚠ TO CONFIRM
  • Intended use: observation encoding paired with the DASF policy above; not intended to be used with any other policy.
  • Out-of-scope use: as above.
  • Known limitations: ⚠ TO CONFIRM
  • Redistribution status: ⚠ TO CONFIRM

Update procedure

Any change to a *.onnx file under controllers/model/ must, in the same pull request:

  1. Update the size in the "Model index" table above.
  2. Update the checkpoint id / hash and training-run fields in the corresponding section.
  3. Update the matching row in THIRD_PARTY_NOTICES.md §2.
  4. Update CHANGELOG.md under [Unreleased].

CI enforces (1) and (3): a *.onnx addition without a matching THIRD_PARTY_NOTICES.md and MODEL_CARD.md reference will fail the merge.