Issues

• Learning Rate for Real World robot finetuning

  OpenDriveLab/UniVLA#84 · Filter by author Jkho80 opened on Sep 27, 2026

• Verify evals on Papers with Code

  OpenDriveLab/UniVLA#81 · Filter by author NielsRogge opened on Jul 17, 2026
  1

• Higher memory usage on GPU0 than other GPUs when finetuning univla-7b on Libero datasets

  OpenDriveLab/UniVLA#79 · Filter by author PuzhenYuan opened on Mar 31, 2026

• ENV problem

  OpenDriveLab/UniVLA#78 · Filter by author DingdongD opened on Jan 24, 2026

• stage 1 loss

  OpenDriveLab/UniVLA#77 · Filter by author waynechu1021 opened on Jan 18, 2026
  2

• Regarding deployment on a real robotic arm

  OpenDriveLab/UniVLA#76 · Filter by author zdq2002 opened on Jan 13, 2026

• self.dino_encoder = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitb14_reg')

  OpenDriveLab/UniVLA#75 · Filter by author 2561681244 opened on Jan 9, 2026

• Question about calvin inference

  OpenDriveLab/UniVLA#73 · Filter by author Hitjss opened on Dec 24, 2025
  2

• Possible mislabeling in Fig. 8: LIBERO example in Group B appears to depict a pick-up action rather than a place action

  OpenDriveLab/UniVLA#69 · Filter by author senp98 opened on Nov 17, 2025

• A question about the process of action in BridgedataV2

  OpenDriveLab/UniVLA#68 · Filter by author xianzeYao opened on Oct 28, 2025

• Question about finetune code for qwbu/univla-7b-224-sft-simpler-bridge

  OpenDriveLab/UniVLA#67 · Filter by author minghaoguo20 opened on Oct 24, 2025

• About exploiting wrist view camera

  OpenDriveLab/UniVLA#65 · Filter by author Eddiesyn opened on Sep 24, 2025
  1

• Finetuning on LIBERO using pretrained checkpoint with bridgev2-only dataset

  OpenDriveLab/UniVLA#64 · Filter by author feiyu12138 opened on Sep 11, 2025

• Loss instability when resuming pretraining from converted HF weights

  OpenDriveLab/UniVLA#60 · Filter by author minghaoguo20 opened on Aug 20, 2025
  11

• training metric train/code_usage_step

  OpenDriveLab/UniVLA#58 · Filter by author choochooyaya6-blip opened on Aug 18, 2025
  3

• menory leak

  OpenDriveLab/UniVLA#57 · Filter by author JackHuang0701 opened on Aug 16, 2025
  4

• Question about potential inconsistency in latent_action selection between training and evaluation

  OpenDriveLab/UniVLA#54 · Filter by author shiyouheng opened on Aug 10, 2025
  4

• The issue of the time ratio between pre-training and fine-tuning

  OpenDriveLab/UniVLA#52 · Filter by author dajiebi opened on Aug 7, 2025
  2

• load checkpoint for stage1 of lam and only training stage2, but the loss becoming "Nan" after a few update

  OpenDriveLab/UniVLA#40 · Filter by author iFLYTEK-Embodied-Robotics-Team opened on Jul 15, 2025
  2

• How to get the attention heatmap in Fig. A-3?

  OpenDriveLab/UniVLA#36 · Filter by author B0B8K1ng opened on Jul 9, 2025
  4

• Request for sharing training logs

  OpenDriveLab/UniVLA#33 · Filter by author dongdongunique opened on Jul 4, 2025
  2

• Question about LAM stage 2 didn't use text encoder

  good first issue
  OpenDriveLab/UniVLA#14 · Filter by author lucasjinreal opened on May 25, 2025
  7

• Online inference script of LAM model

  OpenDriveLab/UniVLA#13 · Filter by author lucasjinreal opened on May 25, 2025
  7

• how to maxmize GPU utils?

  OpenDriveLab/UniVLA#11 · Filter by author Luo-Zhongwei opened on May 22, 2025