Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions docs/downloads/reports/DVIDIA Observation Lab pilot.md
Original file line number Diff line number Diff line change
Expand Up @@ -62,3 +62,25 @@ The frozen authored benchmark includes stationary motion, constant velocity, sto
The batch ran once on macOS ARM64 with Python 3.12.9, OpenCV 5.0.0 and NumPy 2.5.3. Processing covers intake, probing, streaming decode, tracking/forecast, writes and scoring; the offline processing/source-duration ratio is 0.0606. One OpenCV thread was requested while its runtime reported ten; actual concurrency and an enforced thread limit remain unmeasured. Fast replay does not establish live capture-to-command delay or a reliable 30 Hz robot loop. The 250 ms forecast falls between source frames and uses a disclosed nearest-frame reference, so it is not an exact-time accuracy result. Heuristic uncertainty envelopes remain uncalibrated. ([Measured run and limitations](https://github.com/Dvidia-Inference/dvidia-training/blob/ade2b541d8ca4a785bbd1effdfd33eb2a306f150/benchmarks/motion-alpha-20261010/README.md))

Local validation recorded **378 repository tests and six browser workflow checks** for this candidate. These are software/interface checks; no CI completion count, independent real-motion accuracy, live sensing or physical control result follows from them. Original DVIDIA motion, fixture and scoring code uses MIT, while existing upstream detector notices retain their separate terms. Next evidence is independently reviewed permitted motion footage, measured observation-only live streaming and calibrated geometry, followed by synchronized robot actions/state for interaction-conditioned prediction. ([Reviewable candidate and validation](https://github.com/Dvidia-Inference/dvidia-training/pull/7), [motion and interaction gates](https://github.com/Dvidia-Inference/dvidia-training/blob/ade2b541d8ca4a785bbd1effdfd33eb2a306f150/docs/motion-prediction-plan.md))

## Compact CPU pilot learns the next actuator boundary

On 11 October 2026, a separate laptop-first source pilot learns **seven next-20-ms actuator deltas** from native episode tapes. Its 20 inputs are current joint positions, joint velocities and measured jaw gap, plus issued joint and jaw targets. Weighted ridge regression fits on training groups; development groups choose regularization. The selected model is saved before test-tape parsing and transition-target extraction. Authored phases, intervention names, configuration, future state, object/contact truth and final applied servo controls do not enter its feature vector. The measurements are privileged simulator state in sensor-shaped fields, not calibrated physical encoders or observations recovered from human footage. ([Exact source commit 0d3f05f](https://github.com/Dvidia-Inference/dvidia-training/tree/0d3f05f186423e18b15ed8362d772b7d74ec0207), [feature contract and runnable commands](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/docs/laptop-skill-pilot.md), [episode evidence](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/docs/episode-tape.md))

The corpus records **36 attempts across 12 seeded native scene groups**, each with nominal, open-jaw and release-after-lift commands. All sibling interventions stay together: eight groups train, two select on development and two form the test partition. The authored rigid-box controller generates every recording, with predeclared scene jitter and bounded mass/friction variation. All recordings complete; 24 terminal partial boundaries remain explicit exclusions. The test scores **4,553 retained transitions from six recordings**, not 4,553 independent robot trials. The recording controller's 24 simulated successes and 12 horizon failures are not learned-model success rates. Manifest outcomes are available during preflight; the freeze declaration concerns unread test transition targets specifically. ([Corpus and denominators](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/README.md), [frozen model selection](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/freeze.json))

| Test predictor | Joint RMSE (rad) | Jaw RMSE (mm) | Jaw MAE (mm) | Worst jaw error (mm) |
| --- | ---: | ---: | ---: | ---: |
| Action-conditioned ridge | 0.000036306 | 0.341609 | 0.114485 | 3.424593 |
| State-only ridge | 0.000410349 | 0.551878 | 0.188971 | 3.551034 |
| Persistence | 0.002607246 | 0.564418 | 0.108018 | 3.200040 |

The action model improves joint and jaw RMSE over these baselines, but its **jaw MAE and worst error regress against persistence**. Constant joint velocity and training-mean delta are also retained in the full results; constant velocity predicts zero jaw delta because jaw velocity is not an input. On the nearly stationary open-jaw test case, action-model jaw RMSE is 0.057758 mm against 0.002761 mm for persistence. No mixed-unit accuracy percentage or universally better actuator behavior follows from the aggregate scores. ([Exact final report and case/slice results](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/report.json))

An identical-state nominal/release pair around 5.8 simulation seconds exposes a larger transient problem. Both have the same before-state and joint request, while the jaw target changes from 33.538 mm to 80 mm. The action model predicts a 3.430 mm next-boundary opening, but the native observation opens only 0.0104 mm. The state-only model gives the same answer for both branches. This establishes command sensitivity while revealing a serious opening overprediction; it does not validate contact dynamics or qualify an action-conditioned world model. ([Matched intervention receipt](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/README.md))

The action artifact contains **6,438 bytes**. On the recorded macOS ARM64 host, 100 warmup calls followed by 1,000 measured calls give **0.207 ms p95 model computation**, with 0.198 ms median and 0.232 ms worst. The timed call includes model validation and arithmetic over pre-extracted test features. It excludes camera capture, perception, feature extraction, transport, actuator response and cold saved-model loading; it is not robot reaction time or an operating-system deadline guarantee. Fitting uses NumPy, while saved-model inference uses Python's standard library. Corpus generation requires the optional native arm dependency; no GPU or model API participates. The process-lifetime memory high-water mark is 176.062 MiB, including earlier work, rather than an isolated learner requirement or minimum laptop specification. Local software validation records 424 passing tests, with no physical accuracy claim. ([Measured runtime boundaries](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/README.md), [open the offline interactive report](https://research.dvidia.org/downloads/topics/physical-grounding/runs/2026-10-11-actuator-pilot/report.html))

This pilot adds a learned command-to-motion component to the Skillspace path. Reviewed footage can supply task structure and visual evidence; synchronized teleoperation must supply actual robot actions and measured state before a policy can learn corresponding movements. A future compact policy needs independent task benchmarks and a compatible robot/gripper adapter. ACT's action-chunking approach is one research candidate, and LeRobot documents conditional CPU checkpoint deployment with a measured control-rate requirement. Neither establishes the speed or shoe competence of a DVIDIA policy. ([ACT research](https://tonyzhaozh.github.io/aloha/), [LeRobot inference documentation](https://huggingface.co/docs/lerobot/main/inference))

The accompanying shoe collection plan begins with **one shoe and one marked slot**, a fixed calibrated camera/table, a verified grasp region and tucked laces. Its proposed 120-attempt, 12-pair workload and per-condition group quotas are study estimates, not known training sufficiency thresholds. Proposed 15-degree orientation, 20-mm position and one-second stability criteria need calibrated observable measurement procedures before validation. Real shoe evidence, physical sensing, a hardware adapter and validated shoe physics remain absent; the shoe benchmark stays blocked. The current predictor issues no commands and learns no shoe policy. Original DVIDIA coefficients and evidence are published under MIT, with no raw tapes, private footage, credentials or third-party pretrained weights in the bundle. The frozen Observation Lab v0.1 release and synthetic Hugging Face review demo remain separate from this source pilot. ([Proposed capture plan](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/examples/shoe-organizing-coverage.json), [pipeline and remaining gates](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/docs/laptop-skill-pilot.md), [artifact license](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/LICENSE))
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
MIT License

Copyright (c) 2026 DVIDIA contributors

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
# Compact CPU actuator pilot, 11 October 2026

[Open the interactive report](report.html). This exact simulation receipt and
original MIT model bundle comes from DVIDIA Training's
[measured source commit](https://github.com/Dvidia-Inference/dvidia-training/tree/0d3f05f186423e18b15ed8362d772b7d74ec0207).
The [complete benchmark](https://github.com/Dvidia-Inference/dvidia-training/blob/0d3f05f186423e18b15ed8362d772b7d74ec0207/benchmarks/actuator-pilot-20261011/README.md)
records baselines, denominators, transient failures and timing boundaries.

All six JSON receipt/model files are exact final-run bytes. `source.json` is a
post-run source reference, and `report.html` is a self-contained reader of the
exact receipt. The source corpus manifest retains relative native tape paths
and hashes; tapes are omitted. No footage, credentials or third-party weights
are distributed here.

With the candidate source installed, `dvidia-transition inspect DIRECTORY`
checks the five bound artifacts against `integrity.json`. This verifies internal
consistency only, not source authenticity or physical accuracy. Original
DVIDIA coefficients and evidence use the supplied [MIT license](LICENSE).

The predictor estimates the next 20 ms of joint/jaw motion from native state
and issued commands. It issues no commands, predicts no task success and is
unqualified for physical robots. Shoe data, shoe physics and a hardware adapter
remain open. Model-only compute timing excludes sensing and actuation.
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
{"alpha":0.0001,"coefficients":[[-0.023413911279235672,0.00013407676756472406,0.0003203766229395209,3.385448333636372e-06,0.00017321301275480667,0.020132376225264318,0.0006013207924994154,4.430150692606901e-07,7.886716196884903e-06,3.4779427934558955e-06,1.803514766483e-06,0.00023824704406615052,5.407164943187078e-07,0.022032663129652248,0.0006466113856412709,0.0010816819321729197,-7.6104386552289184e-06,0.000801525286877072,-0.020931580895070762,2.506335241894005e-07],[0.028897060212908034,-0.03075628405115735,0.018075803960955157,2.7692359613556942e-05,0.011004291463574851,0.028813033274637183,0.0005214717235201334,0.0005487569321548179,0.00042473480840509765,-1.6987682400190732e-05,0.00015540088408312448,0.00024105210598963366,-3.6941762690573377e-06,0.0010797254969632498,0.037538966792016605,-0.0065010137849883885,-1.309353702264536e-05,-0.002867445897627287,0.001017151473169434,-7.55676049845229e-06],[-0.06837042685742636,0.00636055965476416,-0.037459901610317384,5.8408945891694886e-05,0.02154120182035249,-0.06304671798033504,-0.001650886909583609,0.00013064025919859378,0.0021642579818388612,-3.429730369525898e-06,0.0006796452382266949,-0.0008810232082383357,-1.0784527092206725e-05,-0.0022562537849201386,-0.02560019480716129,0.0033559491718872757,-6.563433354691982e-05,-0.045386537137997916,-0.00713874394944704,7.93037189129509e-06],[0.0005810621511276375,-6.609704904846584e-06,-9.555805711105124e-05,-2.8225119969198456e-06,-4.624575739576662e-05,0.00034809326159791147,1.627695604684937e-05,-2.7975119521258152e-08,-3.465369326916786e-06,1.1245225381633713e-06,-1.3938233543710135e-06,5.963562340117197e-06,-8.741771910459407e-08,-8.026010259530823e-05,2.9552246239712374e-05,0.00013758714242834038,2.0800472826216233e-06,7.490805308325208e-05,0.00014652955206523736,3.3369971356306223e-08],[0.04204767706576623,0.004941037866092892,-8.614408755212916e-05,3.533611689534423e-05,-0.0730657021495212,0.040280083782873705,0.0008785599184403115,8.389381136526186e-06,-0.00037785324562798886,1.3038922856525207e-05,0.00025267696419823294,0.00044293823354058057,-6.8952941412634566e-06,0.00039925799818446654,-0.016675829604721082,-0.020783407916172357,-3.270449344164002e-05,0.05846580976024862,0.0019056080571623136,-1.209474061041307e-05],[0.027111811902137423,-1.9314636573421795e-05,-0.0010017447613202527,-1.138432906652123e-05,-0.0005011040107069056,-0.04091116580119544,0.0005658641452908233,2.0310824715503414e-06,-3.4169253489535234e-05,-8.38926729722339e-06,-1.4136569114327172e-05,0.00040498859001560517,-2.782182568318342e-07,-0.033938523136976304,-0.001197370813522306,-0.0011661682083346757,1.1436829824493793e-05,-0.0010107838521480428,0.03321016529468509,4.4133603330231703e-07],[0.26599189884719415,0.052313045664778034,0.08494490416660601,5.187975935088541e-05,0.05556362572906438,0.25770663545560346,0.00528639711577106,0.0013277000745836854,0.002225432003594018,-0.00019222466598737875,0.001107456667416646,0.0025443590514423083,-0.0013900792209215517,-0.004751127625655068,-0.0035234684087305235,0.0024226896146674446,0.00012666435597677748,0.005646375222438801,0.002180006869283331,0.0016143535524403483]],"control_dt_seconds":0.02,"feature_contract":{"acquisition":"required signals contemporaneous with boundary; no delayed-sensor propagation","availability":"before decision plus issued command; no future or applied control","names":["before.joint_position[0]","before.joint_position[1]","before.joint_position[2]","before.joint_position[3]","before.joint_position[4]","before.joint_position[5]","before.joint_velocity[0]","before.joint_velocity[1]","before.joint_velocity[2]","before.joint_velocity[3]","before.joint_velocity[4]","before.joint_velocity[5]","before.gripper_width","command.issued.joint_targets[0]","command.issued.joint_targets[1]","command.issued.joint_targets[2]","command.issued.joint_targets[3]","command.issued.joint_targets[4]","command.issued.joint_targets[5]","command.issued.gripper_width"],"origin":"sensor-shaped privileged native simulation; hardware sensors unqualified","units":["rad","rad","rad","rad","rad","rad","rad/s","rad/s","rad/s","rad/s","rad/s","rad/s","m","rad","rad","rad","rad","rad","rad","m"]},"feature_mode":"action","intercept":[0.0009610293150821192,0.0002868151938799142,1.1362007785179062e-05,1.1868488215456403e-10,-0.00029294613886360224,-0.0009610293636950372,-7.577548372409316e-09],"is_controller":false,"kind":"dvidia.compact-action-transition-model","license":"MIT","model_sha256":"c63cf993c733687a02217e5b060c711c09878e32d0526feeb44abfa0170e29f3","normalization":{"constant_features":[false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false,false],"means":[-0.0899905108472486,-1.1519101433125067,1.3351944186598976,2.0685882094698694e-06,-0.18334919247409012,0.08884156095454729,0.04256845907588669,0.01272885406120272,0.0012493047510298584,0.00011965114235991223,-0.012338512856367419,-0.038178765041471746,0.06767697721595069,-0.08650386989124388,-1.1509821105807483,1.3351471309301313,1.146019022818662e-06,-0.1841626660141949,0.08651220655980689,0.06582770080989356],"scales":[0.17860326146676422,0.1767063627527591,0.31658147968533146,7.945667011746564e-06,0.22167432531872702,0.1777062994467045,0.06463297032762196,0.08931124694281727,0.1797614962369242,0.00022410177231653518,0.15581823686916052,0.05904040544125759,0.018225447281794958,0.1757749114495912,0.17814145082896937,0.31831002701662303,6.997590745910784e-06,0.22144346050485844,0.17578204958614485,0.021410149727423857],"scope":"fit-training-only"},"output_contract":{"acquisition":"required after signals contemporaneous with the after boundary","contact_and_task_outcomes":"not predicted","definition":"after minus before at exactly one 20 ms control boundary","names":["delta.joint_position[0]","delta.joint_position[1]","delta.joint_position[2]","delta.joint_position[3]","delta.joint_position[4]","delta.joint_position[5]","delta.gripper_width"],"units":["rad","rad","rad","rad","rad","rad","m"]},"physical_robot_ready":false,"predicts_task_success":false,"schema_version":1,"scope":"simulation-only","sensor_origin":"privileged-native-simulation","training":{"positive_weight_rows":18228,"rows":18228,"rows_sha256":"7bff06c153c5071b3b162eec08ea23eaeeee032daf7d1fb0861c6d9cfc674248","sample_weights_sha256":"7787fd55dcedecc0ac29b9030d915ca9036afaa6247c5484333ac4dd9bfc8596","weight_sum":18228.0}}

Large diffs are not rendered by default.

Loading
Loading