Kite Trossen Cup Sim Baseline ACT

공식 가중치
논문 및 기술 보고서

Kite’s task-trained LeRobot ACT policy for red-cup picking on a bimanual WidowX AI station. This baseline-only checkpoint uses 17 simulated episodes and 90,000 training steps; evaluation is by the publisher in simulation only.

전체 모델 설명 보기

Research and simulation evaluation of red-cup placement on the specified bimanual WidowX AI station.

논문 및 기술 보고서

논문 / 보고서 미확인

공식 소개 자료가 등록되어 있으며 해당 버전의 논문이나 기술 보고서는 아직 확인되지 않았습니다.

최근 확인: 2026-11-18

접근 및 사용 방법

모델 파일
공개 접근
학습 데이터
공개 접근

평가 자료

이 버전의 학습 코드는 미확인입니다. 공개 코드가 이 버전의 학습 구현을 포함한다는 뜻은 아닙니다. 검증 설명 보기

학습 / 미세 조정 코드

모델 파일

모델 파일
trossen-bin-picking-sim-act-cup-baseline-90k · 90000접근 방법가중치 라이선스
출시 정보

출처확인 날짜 · 2026-11-18

배포 전 확인 사항

Single cup task; author simulation only, no right-layout success or real-robot validation. Full training scripts, action units, control frequency, minimum VRAM and file-level license scope remain unverified. 가중치나 코드가 공개되어도 로봇에서 즉시 실행할 수 있다는 의미는 아닙니다. 문서에서 구성, 데이터 및 인터페이스 요구 사항을 확인하세요.

배포 및 사용 자료

추론 방식
ACTPolicy.from_pretrained; LeRobot >=0.6; pin revision
실행 환경 및 프레임워크
lerobot-pytorch
미세 조정 요구 사항
Task training: 90,000 steps, batch 8, one A100, about 4 hours
배포 요구 사항
Four 3×480×640 RGB cameras and 14 joint positions; bimanual WidowX AI

구조 및 입출력

모델 유형
행동 정책
아키텍처
Transformer
백본
ResNet18 (ImageNet1K V1), Transformer 512, encoder 4 / decoder 1
입력
RGB 이미지 / 로봇 상태
출력
14 joint targets; 100-step chunks
출시일
2026년 11월 17일

적용 범위 및 학습 방식

기반 모델

Hugging Face LeRobot ACT

본체 범위

단일 본체

기능 분류

Action control

기록 유형

공식 가중치

출력

행동

파이프라인 작업

robotics

예측 출력

chunk_size=100; n_action_steps=100

로봇 형태

양팔 플랫폼 / 로봇 팔

작업 범위

단일 작업

학습 데이터 규모

6,867 training frames

학습 데이터 출처

Kite simulated baseline, 17 training cup episodes

학습 방식

모방 학습

계보

파생 모델: ACT · LeRobot ACT ↗

  • 기반 모델 · 본체 범위 · 기능 분류 · 기록 유형 · 출력 · 파이프라인 작업 · 로봇 형태 · 작업 범위 · 학습 데이터 규모 · 학습 데이터 출처 · 학습 방식 · 계보 · LeRobot ACT: Kite ML, Inc.
  • 예측 출력: Kite ML, Inc.

관련 로봇

전체 매개변수 및 검증 설명 (모델 카드)

Kite ML, Inc. made this task-trained derivative of Hugging Face LeRobot ACT. It is not a new foundation algorithm; Trossen Robotics provides the original recordings and robot, not this derivative policy. It is distinct from baseline 30,000 steps, all-seven 90,000 steps and real-recording 30,000 steps. No particular pretrained ACT checkpoint is established as its initialization.

Input: four RGB views (low, high, left wrist, right wrist), each 3×480×640, and 14 joint positions. Output: 14 joint targets in 100-step chunks; n_obs_steps=1, chunk_size=n_action_steps=100. It does not read language instructions. Configuration: ResNet18 with ImageNet1K V1 weights, Transformer width 512, 8 heads, feedforward 3200, ReLU, encoder 4 / decoder 1; VAE latent 32 and 4 VAE encoder layers, dropout 0.1, KL weight 10. MEAN_STD normalization for visual/state/action. Saved settings: CUDA, AMP/PEFT off, learning rate 1e-5, backbone learning rate 1e-5, weight decay 1e-4. These are not minimum inference requirements.

Training: kiteml/trossen-bin-picking-sim-baseline, 17 cup episodes, 6,867 frames; 90,000 steps, batch 8, one A100, about 4 hours, as reported by the author. LeRobot >=0.6 is the stated loader. The pinned experiment repository contains README, videos and results, without full training or rollout scripts; exact upstream training revision and full reproducibility remain unverified. The upstream LeRobot repository is implementation context, not proof that this exact training pipeline is published.

Author-only closed-loop MuJoCo evaluation: room light 8/20, low light 3/20, warm light 5/20, extra objects 10/20; total 26/80 (32.5%). Two held-out layouts from source episodes 4 (left pickup) and 14 (right pickup), 10 starts each per condition, perturbations within 1 cm. Placed means lifted at least 3 cm and finally inside the tray, max 19 s, complete 100-step plans executed. All successes were on the left layout; none on the right. Images are rendered by the simulator; the low view includes a real-station photo and cup textures. Wrist-camera cables now follow dataset shape and arm joints instead of fixed hooks. Evaluation protocols changed, so old/new results cannot directly measure model improvement. No real-robot run, independent reproduction or generalization proof is claimed.

Complete weights opened 2026-11-17 23:38:54 UTC (2026-11-18 07:38:54 Taipei), not the empty repository at 23:38:31 UTC. Public and nongated. model.safetensors: 206,765,304 bytes. We downloaded the complete file and verified SHA256 c2e798686cb00f6ff007285ce3f2907e4a92844f8fbf8fde333d7c55226941b8 against the LFS declaration, without running the policy. Configuration, pre/postprocessing and normalization files are present. Pin revision 353a961fb1c17a2f00b95fcbac4994282faf8eeb when calling ACTPolicy.from_pretrained.

Model-card metadata declares Apache-2.0; config.license is null and the repository has no standalone LICENSE. This is a publisher declaration, not a file-by-file legal audit; commercial permission stays unknown. Action units/control frequency, full-model trainable parameter count and minimum VRAM are unverified.

https://huggingface.co/kiteml/trossen-bin-picking-sim-act-cup-baseline-90k/blob/353a961fb1c17a2f00b95fcbac4994282faf8eeb/README.md

https://huggingface.co/kiteml/trossen-bin-picking-sim-act-cup-baseline-90k/blob/353a961fb1c17a2f00b95fcbac4994282faf8eeb/config.json

https://github.com/kite-ml/trossen-bin-picking-sim-datasets/tree/f61bcdc1211db23873c28aa785cb4ada8ccccaad

출처 또는 수정 제안 보내기