All pull requests

Pull requests

• feat: enable Hugging Face eval_loss on sharded datasets

  #777·DanielAsadi opened Sep 22, 2026·

  ·

• fix(data): clamp history delta_indices at episode start instead of wrapping to the episode tail

  #772·elbourne12345 opened Sep 5, 2026·

  ·
  1

• Remove the dead use_mean_std flag from the N1.7 processor

  #770·rootkiller6788 opened Sep 3, 2026·

  ·

• setup_SimplerEnv: quote the json-numpy version spec

  #766·rakhimovv opened Sep 2, 2026·

  ·

• tests: snapshot visible-GPU count so the multigpu test runs

  #765·rakhimovv opened Sep 2, 2026·

  ·

• tests: skip installed sim virtualenvs in the env-prefix scan

  #764·rakhimovv opened Sep 2, 2026·

  ·

• convert_v3_to_v2: allow dataset paths under root's home

  #763·rakhimovv opened Sep 2, 2026·

  ·

• Keep normalization results floating point for integer inputs

  #758·catplotlib opened Aug 24, 2026·

  ·
  2

• fix(droid): use extrinsic xyz euler convention for eef rot6d

  #757·XingxinHE opened Aug 21, 2026·

  ·
  1

• fix(eval): fail Vulkan readiness setup errors

  #754·sylvesterkaczmarek opened Aug 15, 2026·

  ·

• fix(policy): honor replay action delta indices

  #753·sylvesterkaczmarek opened Aug 15, 2026·

  ·

• fix(config): harden validation error paths

  #752·sylvesterkaczmarek opened Aug 15, 2026·

  ·

• fix(config): isolate default modality configs

  #751·sylvesterkaczmarek opened Aug 15, 2026·

  ·

• Honour processor overrides passed to from_pretrained on the finetune path

  #744·jiajingc-nv opened Aug 10, 2026Contributor·

  ·

• Add: sparse video decoding for shard step indices

  #743·garlic-byte opened Aug 7, 2026·

  ·
  1

• Add opt-in strict action padding masking in GR00T N1.7

  #741·Henry-23 opened Aug 5, 2026·

  ·
  6

• Export action-aligned policy features

  #738·knanxu opened Aug 4, 2026·

  ·

• add: real robot -- rebot-arm-rs

  #737·yuyoujiang opened Jul 31, 2026Contributor·

  ·
  8

• Project cross-attention K/V once per inference instead of once per denoising step

  #736·dvdimitrov13 opened Jul 30, 2026·

  ·

• Skip full-vocabulary logits the backbone never reads (logits_to_keep=1)

  #734·dvdimitrov13 opened Jul 30, 2026·

  ·
  1

• Replace deprecated torch.backends.cuda.sdp_kernel with torch.nn.attention.sdpa_kernel

  #733·dvdimitrov13 opened Jul 30, 2026·

  ·

• Bump setuptools from 82.0.1 to 83.0.0

  dependenciespython:uv
  #731·dependabot[bot] opened Jul 22, 2026Bot·

  ·

• Bump pillow from 12.2.0 to 12.3.0 in /scripts/deployment/thor

  dependenciespython:uv
  #729·dependabot[bot] opened Jul 21, 2026Bot·

  ·

• Expose gradient accumulation in finetune.sh for low-VRAM fine-tuning

  #726·maxwbuckley opened Jul 16, 2026Contributor·

  ·
  3

• Load policy checkpoints directly in bf16 to halve peak load memory

  #725·maxwbuckley opened Jul 16, 2026Contributor·

  ·
  1