Skip to content

Pull requests: radixark/miles

New pull request New
329 Open 1,055 Closed
329 Open 1,055 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Add packed SFT rollout for pretokenized blocks
#1515 opened Jun 29, 2026 by samaritan1998 Loading…
8
feat: Dynamo integration
#1514 opened Jun 29, 2026 by AndyDai-nv • Draft
7
chore: make megatron e2e CaseConfig topology explicit instead of inferred run-ci-megatron
#1513 opened Jun 29, 2026 by guapisolo Collaborator Loading…
3
Megatron e2e: weight-check skip-list, Qwen3.5 MTP cases run-ci-qwen35 Run qwen3.5 e2e CI
#1512 opened Jun 29, 2026 by guapisolo Collaborator Loading…
8
feat(ci): metric-history gate foundation — storage contract + collection backend (M0+M1)
#1511 opened Jun 29, 2026 by guapisolo Collaborator Loading…
7
refactor: extract session core with direct HTTP responses
#1510 opened Jun 29, 2026 by guapisolo Collaborator Loading…
7
[test] mv deepseek v4 to /manual
#1508 opened Jun 29, 2026 by yueming-yuan Collaborator Loading…
Ci history gate
#1507 opened Jun 29, 2026 by guapisolo Collaborator • Draft
7
[AMD] update ROCm sglang base image
#1506 opened Jun 29, 2026 by XinyuJiangCMU Contributor Loading…
3
fsdp: keep fp32 master for nemotron_h (mixed-dtype checkpoint)
#1502 opened Jun 29, 2026 by Zhichenzzz Contributor Loading…
fsdp: clear stale GDN packing boundaries on non-packed forwards
#1501 opened Jun 29, 2026 by Zhichenzzz Contributor Loading…
2
fsdp: force flash attention for attention-sink models (gpt-oss)
#1500 opened Jun 28, 2026 by Zhichenzzz Contributor Loading…
2
feat(loss): support pg_loss aggregation modes
#1498 opened Jun 27, 2026 by EazyReal Loading…
5
fix(update_weight): skip flush_cache for retract pause mode
#1497 opened Jun 27, 2026 by Shi-Dong Contributor Loading…
1 task
fix(opd): score teacher at rollout temperature
#1496 opened Jun 27, 2026 by EazyReal Loading…
4
fix(dist): preserve new_group options across reload
#1495 opened Jun 27, 2026 by EazyReal Loading…
3
fix(train): support eval-only mode (--num-rollout 0)
#1494 opened Jun 27, 2026 by EazyReal Loading…
2
fix(ppo): preserve raw KL metric tensor
#1493 opened Jun 27, 2026 by EazyReal Loading…
fix(weights): handle empty colocated tensor buckets
#1492 opened Jun 27, 2026 by EazyReal Loading…
docs: fix cli-reference defaults and advantage-estimator choices
#1490 opened Jun 26, 2026 by Shi-Dong Contributor Loading…
1 task
1
rocm: disable gradient_accumulation_fusion on gfx950 across e2e tests
#1489 opened Jun 26, 2026 by sreerohi Loading…
2
[OPD] Add Qwen3.5-35B-A3B single-node self-distillation example
#1488 opened Jun 26, 2026 by maocheng23 Contributor Loading…
4
Add OpenEnv example: miles <-> HuggingFace OpenEnv integration
#1487 opened Jun 26, 2026 by Shi-Dong Contributor • Draft
1 of 5 tasks
4
fsdp(scripts): RL scripts for representative dense + MoE models
#1486 opened Jun 26, 2026 by Zhichenzzz Contributor • Draft
5
fix(megatron): propagate recompute config to bridge model so checkpointing engages
#1482 opened Jun 25, 2026 by guapisolo Collaborator Loading…
2
ProTip! Filter pull requests by the default branch with base:main.