Commits

Branch selector

 pytorch

User selector

All users

All time

Commit History

Commits on Mar 27, 2026

• Add average_summaries context manager (#1829)

  QuantuMope

  authored
  de6a42f

  Copy full SHA for de6a42f

• Have global_norm properly handle when input consists of only None (#1832)

  QuantuMope

  authored
  cef267e

  Copy full SHA for cef267e

• Add bootstrap_only for TDLoss (#1831)

  QuantuMope

  authored
  064f490

  Copy full SHA for 064f490

Commits on Mar 16, 2026

• Add td_error_abs to TDLoss (#1828)

  QuantuMope

  authored
  ab592f9

  Copy full SHA for ab592f9

• Add critic_warm_start_iters to ActorCriticLoss (#1827)

  QuantuMope

  authored
  de51b01

  Copy full SHA for de51b01

Commits on Jan 23, 2026

• Do not include non_persistent buffer in state_dict (#1826)

  emailweixu

  authored
  6276d26

  Copy full SHA for 6276d26

Commits on Dec 10, 2025

• [Bug Fix] Make DDP workers set absl logging as default (#1824)

  QuantuMope

  authored
  8a3b37e

  Copy full SHA for 8a3b37e

Commits on Nov 12, 2025

• Support actor model averaging for ActorCriticAlgorithm (#1820)

  Show description for 160c6fa
  emailweixu

  authored
  160c6fa

  Copy full SHA for 160c6fa

• Fix cuda memory allocation issue caused by fused_linear_act.py (#1822)

  Show description for f2c844e
  emailweixu

  authored
  f2c844e

  Copy full SHA for f2c844e

Commits on Oct 29, 2025

• downgrade torch to 2.7.1, tensorboard to 2.15, remove pathos to avoid conflict with openpi & lerobot (#1821)

  Show description for f94f059
  le-horizon

  authored
  f94f059

  Copy full SHA for f94f059

Commits on Oct 24, 2025

• Autoconvert experience to f32 only if enable_amp=True (#1819)

  QuantuMope

  authored
  5c9c344

  Copy full SHA for 5c9c344

Commits on Oct 23, 2025

• Allow multiple alf.define_config calls for nested confs (#1816)

  QuantuMope

  authored
  b71090e

  Copy full SHA for b71090e

Commits on Oct 22, 2025

• update protobuf version for openpi (#1818)

  le-horizon

  authored
  ecbde3e

  Copy full SHA for ecbde3e

Commits on Oct 20, 2025

• add uv package management support to alf, update README (#1815)

  Show description for 1eb3f07
  le-horizon

  authored
  1eb3f07

  Copy full SHA for 1eb3f07

Commits on Oct 17, 2025

• Add bnb Adam8bit and AdamW8bit optimizers (#1814)

  Show description for 26fe055
  QuantuMope

  authored
  26fe055

  Copy full SHA for 26fe055

• [tiny] default DistributedUnroller.episode_length to 0 to by default use env's step_type to send data to distributed trainer (#1809)

  le-horizon

  authored
  5259df2

  Copy full SHA for 5259df2

Commits on Oct 10, 2025

• TrainerConfig.save_checkpoint_upon_crash (#1813)

  emailweixu

  authored
  4963ee8

  Copy full SHA for 4963ee8

• Add CustomScheduler (#1812)

  emailweixu

  authored
  c2067f0

  Copy full SHA for c2067f0

Commits on Oct 9, 2025

• AveragedModel utilities (#1811)

  Show description for 5be914c
  emailweixu

  authored
  5be914c

  Copy full SHA for 5be914c

Commits on Oct 4, 2025

• Add entropy field to ActorCriticInfo and PPOInfo (#1808)

  QuantuMope

  authored
  a74759c

  Copy full SHA for a74759c

Commits on Oct 3, 2025

• forward spawn context to create shared replay buffer (#1807)

  Show description for c1c3c5c
  le-horizon

  authored
  c1c3c5c

  Copy full SHA for c1c3c5c

Commits on Sep 27, 2025

• [small] add preprocess_experience_for_replay to Algorithm and call in distributed_off_policy_alg to support overwriting env.step_type reward discount etc. with algorithm generated counterpoarts fro…

  Show description for 0d4efdd
  le-horizon

  authored
  0d4efdd

  Copy full SHA for 0d4efdd

Commits on Sep 19, 2025

• Fix async_eval error "pidfd_getfd: Operation not permitted" (#1805)

  Show description for 29a5332
  emailweixu

  authored
  29a5332

  Copy full SHA for 29a5332

• use ddp for unroll (#1804)

  Show description for 53ab610
  emailweixu

  authored
  53ab610

  Copy full SHA for 53ab610

Commits on Sep 4, 2025

• add http_server:8080/render; fix capture frame; add more logging (#1803)

  Show description for bf8b83f
  le-horizon

  authored
  bf8b83f

  Copy full SHA for bf8b83f

• Fix the out dated CI status badge (#1801)

  Show description for 06ee760
  

  Haichao-Zhang

  and

  Haichao Zhang

  authored
  06ee760

  Copy full SHA for 06ee760

Commits on Aug 28, 2025

• allow alf.bin.play num_steps, due to RL in real setting has no episode end (#1802)

  Show description for d12a517
  le-horizon

  authored
  d12a517

  Copy full SHA for d12a517

Commits on Aug 20, 2025

• Update contributing.rst (#1798)

  Show description for 3e8e394
  le-horizon

  authored
  3e8e394

  Copy full SHA for 3e8e394

Commits on Aug 18, 2025

• handle mask_out_loss_for_last_step=False in RLAlgorithm (#1800)

  Show description for 253d23b
  sqchai

  authored
  253d23b

  Copy full SHA for 253d23b

Commits on Jul 29, 2025

• skip adding param group when no param (#1797)

  Show description for 71a1aa2
  hnyu

  authored
  71a1aa2

  Copy full SHA for 71a1aa2

Commits on Jul 23, 2025

• Summarize the first importance ratio for PPO (#1795)

  Show description for ac142f0
  emailweixu

  authored
  ac142f0

  Copy full SHA for ac142f0

• Add py_map_structure_up_to_with_path (#1796)

  QuantuMope

  authored
  0682417

  Copy full SHA for 0682417

Commits on Jul 16, 2025

• Remove explicit strict checkpoint loading in alf.bin.play (#1794)

  QuantuMope

  authored
  d737776

  Copy full SHA for d737776

• Add TrainerConfig.offline_loss_weight (#1791)

  QuantuMope

  authored
  c40bcd3

  Copy full SHA for c40bcd3

Commits on Jul 9, 2025

• More device support for metrics (#1787)

  emailweixu

  authored
  5f13499

  Copy full SHA for 5f13499