All pull requests

Pull requests

• ci: Bump MBridge.

  CI:L1
  #4752·cspades opened Oct 15, 2026Contributor·4 tasks·

  ·
  4

• fix(nemo-gym): read env.nemo_gym.spinup_timeout for sharded startup

  community-requestDocumentation
  #4751·Thanatos9404 opened Oct 15, 2026·3 of 4 tasks·

  ·
  1

  1

• feat(grpo): add adaptive-ppo-clip

  Documentation
  #4750·smahdavi4 opened Oct 15, 2026Contributor·4 tasks done·

  ·
  1

• fix(nemo-gym): keep unserved-call ledger reasons off the receipt poison list

  CI:L0
  #4749·alexandery-nvidia opened Oct 15, 2026Contributor·

  ·
  3

• feat(speechlm): add native Automodel GSPO ASR recipe

  CI:L2Documentation
  #4748·vmendelev opened Oct 15, 2026·3 of 4 tasks·

  ·
  2

• feat(grpo): length-aware reweighting of positive advantages

  Documentation
  #4747·sdevare-nv opened Oct 15, 2026Contributor·

  ·
  2

• feat(data-plane): mooncake_cpu SSD offload on storage units

  #4746·ZhiyuLi-Nvidia opened Oct 15, 2026Contributor·4 tasks done·

  ·
  1

• chore(gym): declare legacy environment servers in RL configs

  CI:Lfast
  #4745·zyzhou5 opened Oct 15, 2026Contributor·

  ·
  5

• feat(checkpoint): opt-in MSC (S3) checkpoint storage

  CI:L1DocumentationRun CICD
  #4743·ZhiyuLi-Nvidia opened Oct 15, 2026Contributor·

  ·
  7

• cp: cherry-pick #4703 - test(automodel): rename the dtensor-named tests

  CI:L1Documentation
  #4742·yuki-97 opened Oct 15, 2026Contributor·

  ·
  2

• cp: docs: make NeMo Gym the primary evaluation path (4687) into main

  cherry-pickCI:L1Documentation
  #4741·svcnvidia-nemo-ci opened Oct 15, 2026Contributor·

  ·
  2

• test(ppo): add Qwen3-30B-A3B PPO + router replay single-controller nightly

  CI:Lfast
  #4739·bg51717 opened Oct 15, 2026Member·3 of 4 tasks·

  ·
  2

• feat(telemetry): add executor and event-loop diagnostics

  CI:L0Documentation
  #4732·macandro96 opened Oct 14, 2026Contributor·4 tasks done·

  ·
  2

• perf: add a bounded queued vLLM HTTP bridge

  Documentation
  #4731·arnavkomaragiri opened Oct 14, 2026·3 of 5 tasks·

  ·
  1

• feat(telemetry): export rollout checkpoint metrics to Lens

  CI:L0Documentation
  #4730·macandro96 opened Oct 14, 2026Contributor·4 tasks done·

  ·
  15

• fix(checkpoint): stop training at the next step after a failed background save

  community-request
  #4729·Thanatos9404 opened Oct 14, 2026·2 of 4 tasks·

  ·
  1

  1

• feat(xtoken): add opt-in overlength rejection and full-turn CE to collator

  CI:L0Documentation
  #4728·avenkateshha opened Oct 14, 2026Contributor·

  ·
  10

• perf(dataplane): add opt-in native async TQ path for A/B tests

  Documentation
  #4727·zyzhou5 opened Oct 14, 2026Contributor·4 tasks·

  ·
  1

• feat(dynamo): refit, FP8 and speculative decoding via the vLLM worker extension

  CI:L1CIDocumentation
  #4725·jthomson04 opened Oct 14, 2026Contributor·4 tasks done·

  ·
  19

• feat(grpo): add deterministic async rollout scheduling

  CI:L2Documentation
  #4723·wdykas opened Oct 14, 2026Contributor·4 tasks·

  ·
  9

• refactor(xtoken): move same-tokenizer KD math into shared helpers

  Documentation
  #4721·avenkateshha opened Oct 14, 2026Contributor·3 of 4 tasks·

  ·
  6

• fix: fence packed refit buffer reuse across all CUDA streams

  CI:Lfast
  #4720·yuekaizhang opened Oct 14, 2026Contributor·

  ·
  9

• feat(dynamo): port to ai-dynamo 1.5.1 and run under the SingleController

  CI:L1Documentation
  #4719·jthomson04 opened Oct 14, 2026Contributor·4 tasks done·

  ·
  31

• feat: train full rollout trees with GRPO

  Documentation
  #4718·arnavkomaragiri opened Oct 14, 2026·4 of 5 tasks·

  ·
  4

• perf: add direct vLLM token capture for Gym rollouts

  Documentation
  #4717·arnavkomaragiri opened Oct 14, 2026·5 of 6 tasks·

  ·
  4